From d.vanbarneveld at STUDENT.SCIENCE.RU.NL  Thu Jul  6 16:58:40 2006
From: d.vanbarneveld at STUDENT.SCIENCE.RU.NL (Denise van Barneveld)
Date: Thu, 6 Jul 2006 16:58:40 +0200
Subject: problem with text in topoplot and triplot
Message-ID: <THU.6.JUL.2006.165840.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear all,

since I started using the new version of Matlab (R2006a) on a mac, I
have this problem with text in the figures which I generate with
topoplot.m and triplot.m. The upper half of the text lines (title,
colorbar, electrodes, etc) is not visible (see attached figure).
Does anyone have this problem too? Especially people who are using
this new version on a pc. And does anyone perhaps has a solution for
this problem? I am currently looking in the code, if I can find where
it goes wrong, but I wonder whether it is a problem only with the new
version, or only at a mac, or only at my computer.

Best regards,
Denise van Barneveld

-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: application/applefile
Size: 48798 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: image/jpeg
Size: 85086 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment.jpg>

From muthuraman10 at HOTMAIL.COM  Fri Jul  7 12:04:26 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Fri, 7 Jul 2006 10:04:26 +0000
Subject: query on confidence limit, DICS
In-Reply-To: <37FAA4DD-5DD8-4F2F-ABA6-D9F47F6AB95E@fcdonders.ru.nl>
Message-ID: <FRI.7.JUL.2006.100426.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

As i am able to sucessfully use the coherence analysis using freq_mtmfft and
frq_mtmwelch to calculate the coherence
i would like to know is there a way to calculate the confidence limit for
the coherence calculated from the freqanalysis, is there a standard code

Next, i would like to follow with the source analysis 'DICS' is there a
tutorial available

thanking you

with regards
Muthuraman


>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on CTF software!!!
>Date: Fri, 23 Jun 2006 15:23:22 +0200
>
>dear Muthuraman
>
>On 23 Jun 2006, at 12:54, Muthuraman Muthuraman wrote:
>>For the Beamformer source analysis, do we need the CTF software
>>because when i am trying out the tutorial in beamforming
>
>You do not need the ctf software to follow the tutorial. The tutorial  is
>based however on a ctf MEG dataset.
>
>>CTF_READ_RES4 [v  1.12]
>>....done (  0.67 sec)
>>
>>i get the error
>>
>>??? Error using ==> read_fcdc_header
>>could not read CTF res4 header file
>>
>>Error in ==> preprocessing at 263
>>hdr = read_fcdc_header(cfg.headerfile);
>
>Based on your error I suspect that matlab/fieldtrip cannot find the
>dataset that you are trying to analyze, or that you matlab path is
>incorrect. The function "CTF_READ_RES4" is not supposed to be called  (it
>is part of the NIH ctf import routines, e.g. used in EEGLAB),  since the
>default private/read_ctf_res4.m function can deal with the  tutorial
>dataset just fine. Please look into the code of  read_fcdc_header (using
>the matlab debugger) to figure out what your  exact problem is.
>
>best
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul  7 12:36:02 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 7 Jul 2006 12:36:02 +0200
Subject: query on confidence limit, DICS
In-Reply-To: <BAY121-F19F77F86077022EC6849B3CE740@phx.gbl>
Message-ID: <FRI.7.JUL.2006.123602.0200.FIELDTRIP@NIC.SURFNET.NL>

On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
> As i am able to sucessfully use the coherence analysis using
> freq_mtmfft and frq_mtmwelch to calculate the coherence
> i would like to know is there a way to calculate the confidence
> limit for the coherence calculated from the freqanalysis, is there
> a standard code

You can compute a jackknife estimate of the variance in the coherence
using the freqdescriptives function. That requires keeptrials=yes in
freqanalysis.

> Next, i would like to follow with the source analysis 'DICS' is
> there a tutorial available

No there is no tutorial available specifically for that. You can look
at the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
options are cfg.method=dics, cfg.refchan and/or cfg.refdip.

Robert


From j.poort at NIN.KNAW.NL  Tue Jul 11 11:43:49 2006
From: j.poort at NIN.KNAW.NL (Jasper Poort)
Date: Tue, 11 Jul 2006 11:43:49 +0200
Subject: Confidence intervals for evoked potentials
Message-ID: <TUE.11.JUL.2006.114349.0200.FIELDTRIP@NIC.SURFNET.NL>

 
Hi all, I have some questions regarding confidence intervals for timelocked data in FieldTrip. I noted that a field .var is returned by timelockanalysis

line 510 timelockanalysis
% compute the variance
tmp1 = repmat(dofvec(:)', [nchan 1]);
tmp2 = repmat(dofvec(:)', [nchan 1])-1;
var = (ss - (s.^2)./tmp1) ./ tmp2;

Assuming that the sampling distribution is normal I can presumably use this field for computation of confidence intervals of the .avg field? 
Is nonparametric estimation of variance using a bootstrap or jackknife also implemented somewhere? In general I'm just curious to know how others compute confidence intervals for evoked potentials in FieldTrip, for example to get an idea at what point in time two conditions start to differ, 

best regards, Jasper
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060711/806a2020/attachment.html>

From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 15:50:09 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 15:50:09 +0200
Subject: Confidence intervals for evoked potentials
In-Reply-To: <017d01c6a4ce$835853e0$910b57c0@ABSYS>
Message-ID: <TUE.11.JUL.2006.155009.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Jasper,

On 11 Jul 2006, at 11:43, Jasper Poort wrote:
> Hi all, I have some questions regarding confidence intervals for
> timelocked data in FieldTrip. I noted that a field .var is returned
> by timelockanalysis
>
> line 510 timelockanalysis
> % compute the variance
> tmp1 = repmat(dofvec(:)', [nchan 1]);
> tmp2 = repmat(dofvec(:)', [nchan 1])-1;
> var = (ss - (s.^2)./tmp1) ./ tmp2;
>
> Assuming that the sampling distribution is normal I can presumably
> use this field for computation of confidence intervals of the .avg
> field?

Yes, you can use it to compute the channel-timepoint specific
standard-error of mean (SEM).

> Is nonparametric estimation of variance using a bootstrap or
> jackknife also implemented somewhere? In general I'm just curious
> to know how others compute confidence intervals for evoked
> potentials in FieldTrip, for example to get an idea at what point
> in time two conditions start to differ,

Bootstrap and jackknife estimates of variance are only different from
teh normal variance estimate if the measure involves a nonlinear or
biased estimate (i.e. coherence). The jackknife estimate of variance
of the ERP is exactly the same as the normal estimate of variance,
since the ERP is a plain (linear) average. The same holds for
bootstrap estimates.

So that means that for ERPs the jackknife and bootstrap only would be
a computationally expensive methods, resulting in the same variance
as usual. Therefore they are not implemented.

best regards,
Robert


From muthuraman10 at HOTMAIL.COM  Tue Jul 11 16:59:58 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 11 Jul 2006 14:59:58 +0000
Subject: DICS
In-Reply-To: <E2FC4366-C6DD-476C-B902-6EBF864D6B48@fcdonders.ru.nl>
Message-ID: <TUE.11.JUL.2006.145958.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thanks for the suggestion to use the jackknife variance estimate it works
thanks a lot

To use the DICS method

We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG channels
recording
and also we have the 64 channel measurement in FRMI

the electrode locations we have it in a file with 3 columns x,y,z for each
electrode
using the polhemus sensors but our own software

In the Source analysis
Cfg.method='DICS';
but Cfg.xgrid, ygrid and zgrid
the grid values are the sensor location values

or it should be the
cfg.grid.pos-where i can give in the 3 values [x y z] for each electrode

I am a starter in the source analysis in combining EEG-FMRI
please let me know

with regards
muthuraman.






>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on confidence limit, DICS
>Date: Fri, 7 Jul 2006 12:36:02 +0200
>
>On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
>>As i am able to sucessfully use the coherence analysis using  freq_mtmfft
>>and frq_mtmwelch to calculate the coherence
>>i would like to know is there a way to calculate the confidence  limit for
>>the coherence calculated from the freqanalysis, is there  a standard code
>
>You can compute a jackknife estimate of the variance in the coherence
>using the freqdescriptives function. That requires keeptrials=yes in
>freqanalysis.
>
>>Next, i would like to follow with the source analysis 'DICS' is  there a
>>tutorial available
>
>No there is no tutorial available specifically for that. You can look  at
>the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
>doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
>options are cfg.method=dics, cfg.refchan and/or cfg.refdip.
>
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 17:51:28 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 17:51:28 +0200
Subject: DICS
In-Reply-To: <BAY121-F26D094D1C667F7EAEC07B0CE680@phx.gbl>
Message-ID: <TUE.11.JUL.2006.175128.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman

On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
> To use the DICS method
>
> We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG
> channels recording
> and also we have the 64 channel measurement in FRMI
>
> the electrode locations we have it in a file with 3 columns x,y,z
> for each electrode
> using the polhemus sensors but our own software
>
> In the Source analysis
> Cfg.method='DICS';
> but Cfg.xgrid, ygrid and zgrid
> the grid values are the sensor location values
>
> or it should be the
> cfg.grid.pos-where i can give in the 3 values [x y z] for each
> electrode

in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the
locations at which the _dipoles_ will be placed, it is not the
position of the electrodes. At each grid location a spatial filter is
constructed and the power (and coherence) is computed. If you do not
want to scan on a full 3-D grid, but on only a few dipole locations
of interest, you can use cfg.grid.pos to specify those dipole locations.

The electrode positions are specified in the cfg.elec structure, or
in the cfg.elecfile filename. You can find more details on the
definition of electrode positions at http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
_magnetometers_or_gradiometers_described

Furthermore, you can try using the READ_FCDC_ELEC function for
reading the electrode locations, but probably taht will not work
since youy are using your own software for digitization.
Alternatively, you can manually construct an "elec" electrode
definition.

best regards
Robert

PS please upgrade to a recent fieldtrip version, the sourceanalysis
options cfg.xgrid/ygrid/zgrid have been renamed. The old options
still will work, but in general it is preferable to work with a
recent copy of the toolbox.


From muthuraman10 at HOTMAIL.COM  Thu Jul 13 13:11:35 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Thu, 13 Jul 2006 11:11:35 +0000
Subject: DICS
In-Reply-To: <43B3E5A4-716B-48A0-8CE0-5F8810DC4A65@fcdonders.ru.nl>
Message-ID: <THU.13.JUL.2006.111135.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for the suggestions
I have dowloaded the latest version now and i am working with it
i have created the own elec file which is "elec1" and these are the steps i
follow
please go through it and reply me, why is the error

-First i calculate the coherence using the freqanalysis and freqdescriptives
which works perfectly
-Second i calculate the Grid using Prepare_leadfield

vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1);

which also works i get the output for the Grid

-Third I want use the source analysis for the DICS

cfg.method='dics';
cfg.grid=grid;
cfg.elecfile=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
??? Error using ==> fprintf
Function is not defined for 'struct' inputs.

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
  fprintf('reading electrodes from file %s\n', cfg.elecfile);

Error in ==> sourceanalysis at 574
[vol, sens, cfg] = prepare_vol_sens(cfg, data);




-I also tried grid using the freqmtmfft-which as the results for the
freqanalysis
vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1,freqmtmfft);

I get error

using headmodel specified in the configuration
using electrodes specified in the configuration
??? Error using ==> delaunayn
Not enough unique points to do tessellation.

Error in ==> delaunay at 49
    tri = delaunayn([x(:) y(:)]);

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));

Error in ==> prepare_leadfield at 179
[vol, sens, cfg] = prepare_vol_sens(cfg, data);


Thanking you

with regards
Muthuraman.




>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Tue, 11 Jul 2006 17:51:28 +0200
>
>Dear Muthuraman
>
>On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
>>To use the DICS method
>>
>>We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG  channels
>>recording
>>and also we have the 64 channel measurement in FRMI
>>
>>the electrode locations we have it in a file with 3 columns x,y,z  for
>>each electrode
>>using the polhemus sensors but our own software
>>
>>In the Source analysis
>>Cfg.method='DICS';
>>but Cfg.xgrid, ygrid and zgrid
>>the grid values are the sensor location values
>>
>>or it should be the
>>cfg.grid.pos-where i can give in the 3 values [x y z] for each  electrode
>
>in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the  locations at
>which the _dipoles_ will be placed, it is not the  position of the
>electrodes. At each grid location a spatial filter is  constructed and the
>power (and coherence) is computed. If you do not  want to scan on a full
>3-D grid, but on only a few dipole locations  of interest, you can use
>cfg.grid.pos to specify those dipole locations.
>
>The electrode positions are specified in the cfg.elec structure, or  in the
>cfg.elecfile filename. You can find more details on the  definition of
>electrode positions at http://www2.ru.nl/fcdonders/ fieldtrip/doku.php?
>id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
>_magnetometers_or_gradiometers_described
>
>Furthermore, you can try using the READ_FCDC_ELEC function for  reading the
>electrode locations, but probably taht will not work  since youy are using
>your own software for digitization.  Alternatively, you can manually
>construct an "elec" electrode  definition.
>
>best regards
>Robert
>
>PS please upgrade to a recent fieldtrip version, the sourceanalysis
>options cfg.xgrid/ygrid/zgrid have been renamed. The old options  still
>will work, but in general it is preferable to work with a  recent copy of
>the toolbox.


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 14 11:37:47 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 14 Jul 2006 11:37:47 +0200
Subject: DICS
In-Reply-To: <BAY121-F32B67944419449CAB9EEF6CE6E0@phx.gbl>
Message-ID: <FRI.14.JUL.2006.113747.0200.FIELDTRIP@NIC.SURFNET.NL>

On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:

> cfg.method='dics';
> cfg.grid=grid;
> cfg.elecfile=elec1;
> cfg.vol=vol;
> [source]=sourceanalysis(cfg,freqmtmfft);
> using headmodel specified in the configuration
> ??? Error using ==> fprintf
> Function is not defined for 'struct' inputs.
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>  fprintf('reading electrodes from file %s\n', cfg.elecfile);

You should specify either cfg.elecfile with a string (filename) or
cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)

> I get error
>
> using headmodel specified in the configuration
> using electrodes specified in the configuration
> ??? Error using ==> delaunayn
> Not enough unique points to do tessellation.
>
> Error in ==> delaunay at 49
>    tri = delaunayn([x(:) y(:)]);
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>
> Error in ==> prepare_leadfield at 179
> [vol, sens, cfg] = prepare_vol_sens(cfg, data);

That indicates that less than 3 electrodes were selected. Probably
there is a mismatch between the channel names in the data and the
names of the electrodes. FT selects the channels and electrodes based
on the corresponding labels, if they do not correspond then no
channels+electrodes are selected.

Robert


From muthuraman10 at HOTMAIL.COM  Mon Jul 17 16:12:14 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Mon, 17 Jul 2006 14:12:14 +0000
Subject: DICS
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <MON.17.JUL.2006.141214.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
but not when using the mtmwelch is this a problem in the mtmwelch
or i am doing something wrong

Please go through it

thanking you

with regards
muthuraman



>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Fri, 14 Jul 2006 11:37:47 +0200
>
>On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:
>
>>cfg.method='dics';
>>cfg.grid=grid;
>>cfg.elecfile=elec1;
>>cfg.vol=vol;
>>[source]=sourceanalysis(cfg,freqmtmfft);
>>using headmodel specified in the configuration
>>??? Error using ==> fprintf
>>Function is not defined for 'struct' inputs.
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>>  fprintf('reading electrodes from file %s\n', cfg.elecfile);
>
>You should specify either cfg.elecfile with a string (filename) or
>cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)
>
>>I get error
>>
>>using headmodel specified in the configuration
>>using electrodes specified in the configuration
>>??? Error using ==> delaunayn
>>Not enough unique points to do tessellation.
>>
>>Error in ==> delaunay at 49
>>    tri = delaunayn([x(:) y(:)]);
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>>
>>Error in ==> prepare_leadfield at 179
>>[vol, sens, cfg] = prepare_vol_sens(cfg, data);
>
>That indicates that less than 3 electrodes were selected. Probably  there
>is a mismatch between the channel names in the data and the  names of the
>electrodes. FT selects the channels and electrodes based  on the
>corresponding labels, if they do not correspond then no
>channels+electrodes are selected.
>
>Robert


From floris.delange at FCDONDERS.RU.NL  Tue Jul 25 10:24:18 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Tue, 25 Jul 2006 10:24:18 +0200
Subject: sourcestatistics question
Message-ID: <TUE.25.JUL.2006.102418.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear FieldTrippers,

I have a question concerning sourcestatistics.
I have calculated the source localization using DICS for 11 individual
subjects. Now I would like to test for the significance of the sources
using sourcestatistics. The examples I find in the tutorial are about
checking for significance within one subject by shuffling the individual
trials. However, what I would like to do is to do the statistics on the
2nd level (subject), and to test for the consistency of the source
reconstructions across subjects.
Is this possible with sourcestatistics? If so, how should I feed the
data to sourcestatistics? If I say something like stats =
sourcestatistics(cfg,sources{1:11}) where each source{} is a normalized
source (calculated by sourceanalysis, normalized with
sourceinterpolate), it complains (rightfully) that there's a reference
to the non-existing field 'statistic'.

Any ideas are greatly appreciated!

Kind regards,
Floris


From muthuraman10 at HOTMAIL.COM  Tue Jul 25 10:34:30 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 25 Jul 2006 08:34:30 +0000
Subject: Source anaylsis
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.083430.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
i get a error when using the mtmfft for source analysis
cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
using electrodes specified in the configuration
610 dipoles inside, 600 dipoles outside brain

??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
The cross-spectral-density matrix is not complete

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);


Please go through it
Because for the DICS i would like to use the freqmtmwelch
is this a problem with freqmtmwelch
or something wrong in my inputs

thanking you

with regards
muthuraman


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:18:48 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:18:48 +0200
Subject: DICS
In-Reply-To: <BAY121-F235CB8F83C35EABC33B41ECE620@phx.gbl>
Message-ID: <TUE.25.JUL.2006.121848.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:29:23 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:29:23 +0200
Subject: sourcestatistics question
In-Reply-To: <44C5D532.4020708@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.122923.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris,

On 25 Jul 2006, at 10:24, Floris de Lange wrote:
> I have a question concerning sourcestatistics.
> I have calculated the source localization using DICS for 11
> individual subjects. Now I would like to test for the significance
> of the sources using sourcestatistics. The examples I find in the
> tutorial are about checking for significance within one subject by
> shuffling the individual trials. However, what I would like to do
> is to do the statistics on the 2nd level (subject), and to test for
> the consistency of the source reconstructions across subjects.
> Is this possible with sourcestatistics?

Yes, that is possible.

> If so, how should I feed the data to sourcestatistics? If I say
> something like stats = sourcestatistics(cfg,sources{1:11}) where
> each source{} is a normalized source (calculated by sourceanalysis,
> normalized with sourceinterpolate), it complains (rightfully) that
> there's a reference to the non-existing field 'statistic'.


Assuming you want to use a randomization test to get a Monte-Carlo
estimate of the probability of your null-hypothesis
(cfg.method='montecarlo'), cfg.statistic should be a string
describing the statistic whose randomization distribution you want to
compute. See the help of statistic_montecarlo.

Furthermore, the hypothesis that you want to test is about two
conditions. That means that the two conditions both should be fed
into sourcestatistics, e.g. something like

for i=1:11
   read data from condition 1 for each subject
   cond1{i} = sourceanalysis(...)
   cond1{i} = sourceinterpolate(...)
   cond1{i} = volumenormalize(...)
   read data from condition 2 for each subject
   cond2{i} = sourceanalysis(...)
   cond2{i} = sourceinterpolate(...)
   cond2{i} = volumenormalize(...)
end

cfg = []
cfg.design = [1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2    %
condition number
               1 2 3 4 5 .....       1 2 3 4 5 ....       ]   %
subject number
cfg.method = 'montecarlo'
cfg.statistic = 'depsamplesT'
cfg.ivar = 1; % row with the independent variable (condition) in the
design
cfg.uvar = 2; % row with the unit of observation (subject) in the
design matrix
stat = sourcestatistics(cfg, cond1{:}, cond2{:});

It is also possible to do other tests, which are more in line with a
2nd level analysis (e.g. compute t-maps per subject and test over
subjects). It is better to discuss those face-to-face.

best regards,
Robert


From Jan.Schoffelen at FCDONDERS.RU.NL  Tue Jul 25 12:32:38 2006
From: Jan.Schoffelen at FCDONDERS.RU.NL (Jan Mathijs Schoffelen)
Date: Tue, 25 Jul 2006 12:32:38 +0200
Subject: DICS
In-Reply-To: <4442BD57-7F97-469F-990B-CD63E171E371@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.123238.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman,

One little addition to the freq.cumtapcnt-story.
freq.cumtapcnt should be a vector instead of a scalar:
the correct way to specify it would be: freq.cumtapcnt = ones(ntrl,1)
with ntrl the number of trials in your data.

Yours,

Jan-Mathijs

-----Original Message-----
From: FieldTrip discussion list [mailto:FIELDTRIP at NIC.SURFNET.NL] On Behalf
Of Robert Oostenveld
Sent: Tuesday, July 25, 2006 12:19 PM
To: FIELDTRIP at NIC.SURFNET.NL
Subject: Re: [FIELDTRIP] DICS

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:32:52 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:32:52 +0200
Subject: Source anaylsis
In-Reply-To: <BAY121-F324EA9761CB16729FB47E8CE5A0@phx.gbl>
Message-ID: <TUE.25.JUL.2006.123252.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:

> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
> The cross-spectral-density matrix is not complete

this indicates that you did not include all channel combinations
(cfg.channelcmb) in freqanalysis, and that you try to use a cross
spectral density matrix for DISC that is not complete (i.e. come
cross-spectra are missing). You should correct your configuration for
the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
CHANNELCOMBINATION function).

best regards,
Robert


From jciveira at UNAV.ES  Tue Jul 25 14:22:36 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Tue, 25 Jul 2006 14:22:36 +0200
Subject: Source anaylsis
In-Reply-To: <77E33A56-A0D4-43FC-94E0-91A3AC31197B@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.142236.0200.FIELDTRIP@NIC.SURFNET.NL>

I have the same problem and I do not know how to fix it, even though I  
tried your solution with cfg.channelcmb.

My case is : i have a set of 82 electrodes, I make freqanalysis with  
cfg.channelcmb = channelcombination({'all' 'all'},data.label);  thus I  
get

freq =

         label: {82x1 cell}
        dimord: 'chan_freq'
          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000 16.8000  
17.2000 17.6000 18]
     powspctrm: [82x11 double]
      labelcmb: {3321x2 cell}
     crsspctrm: [3321x11 double]
           cfg: [1x1 struct]

When I make sourceanalysis with a dipole as a reference and I have no  
problems but when I want to use a channel as a reference I always have  
the same error:
The cross-spectral-density matrix is not complete.

Can you help me?

Best regards
Juan

Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:
>
>> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
>> The cross-spectral-density matrix is not complete
>
> this indicates that you did not include all channel combinations
> (cfg.channelcmb) in freqanalysis, and that you try to use a cross
> spectral density matrix for DISC that is not complete (i.e. come
> cross-spectra are missing). You should correct your configuration for
> the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
> CHANNELCOMBINATION function).
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From stephan.bickel at ANATOM.UNIZH.CH  Tue Jul 25 17:39:32 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel)
Date: Tue, 25 Jul 2006 17:39:32 +0200
Subject: import bva time-freq data
Message-ID: <TUE.25.JUL.2006.173932.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi,

I try to import brainvision analyzer dat files of wavelet time-frequency
decompositions into fieldtrip. I exported an average file, so it consists
only of a single segment. However I have some problems with it I am not sure
if it is because I define the segment wrong or if it is not possible to import
this data into fieldtrip.
I pasted the code and the error message below and attached the marker- and
header file.

I would be really glad for any suggestions.

Thank you very much,

Stephan 



This is what I tried:

cfg = [];
cfg.headerfile =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vhdr';
cfg.datafile   =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.dat';
cfg.channel = 'all';

%Trialdefinition
cfg.trialdef.trgfile  =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vmrk';
cfg.trialdef.stim  = 'New Segment';
%cfg.trialdef.segment  = 'yes';
cfg.trialdef.timezero = 'yes';
% cfg.trialdef.eventtype  = 'Time 0';
cfg.trialdef.prestim    = 0.500;
cfg.trialdef.poststim   = 1.600;

[cfg] = definetrial(cfg)
[raw] = preprocessing(cfg)


and this is the error message:


evaluating trialfunction 'trialfun_brainvision'
0 stimulus markers converted into 0 trials
0 response markers converted into 0 trials
1 segment markers converted into 0 trials
1 "Time 0" markers converted into 1 trials
found 0 events
created 1 trials

cfg = 

    headerfile: [1x88 char]
      datafile: [1x87 char]
       channel: 'all'
      trialdef: [1x1 struct]
       dataset: []
      trialfun: 'trialfun_brainvision'
         event: []
           trl: [1 1051 -250]
       version: [1x1 struct]

retaining exist trial definition
retaining exist event information
found 0 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060330\private\read_brainvision_eeg at 65
    dat(line,:) = str2num(str);

Error in ==> read_fcdc_data at 419
  dat = read_brainvision_eeg(datafile, hdr, begsample, endsample);

Error in ==> preprocessing at 322
  dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);

-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vmrk
Type: application/octet-stream
Size: 741 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vhdr
Type: application/octet-stream
Size: 3459 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment-0001.obj>

From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:07:03 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:07:03 +0200
Subject: Source anaylsis
In-Reply-To: <20060725142236.lz6o47hwkkwo4sw4@webmail.unav.es>
Message-ID: <WED.26.JUL.2006.090703.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
> I have the same problem and I do not know how to fix it, even
> though I tried your solution with cfg.channelcmb.
>
> My case is : i have a set of 82 electrodes, I make freqanalysis
> with cfg.channelcmb = channelcombination({'all'
> 'all'},data.label);  thus I get
>
> freq =
>
>         label: {82x1 cell}
>        dimord: 'chan_freq'
>          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000
> 16.8000 17.2000 17.6000 18]
>     powspctrm: [82x11 double]
>      labelcmb: {3321x2 cell}
>     crsspctrm: [3321x11 double]
>           cfg: [1x1 struct]
>
> When I make sourceanalysis with a dipole as a reference and I have
> no problems but when I want to use a channel as a reference I
> always have the same error:
> The cross-spectral-density matrix is not complete.
>
> Can you help me?
>
> Best regards
> Juan

Dear Juan

Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
expect the latter, since computing coherence between one particular
EEG electrode and sources in the brain with DICS will not lead to
meaningfull results. The selected EEG electrode will pick up volume-
conducted activity from all sources in the brain (as opposed to an
EMG electrode) and would be referenced to the same reference
electrode as all other electrodes. So there will be a huge amount of
trivial coherence, whose spatial distribution in the brain will
probably more reflect the volume-conduction than something
physiologically interpretable.

Although I have not used DICS with a reference channel on EEG data, I
would expect the code to work if you follow a procedure like this:

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EEG electrodes
cfg.implicitref = name of implicit reference electrode used for the
EEG channels
cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
data_eeg = preprocessing(cfg)

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EMG electrodes, assuming that they are
bipolar channels
% cfg.reref may be used if you have two unipolar channels on the
muscle of interest
cfg.rectify = 'yes'   % this makes the spiky nature of the EMG more
visible, and improves the coherence estimate
cfg.baseline = 'yes'  % remove baseline prior to rectification, not
neccessary but still preferable
data_emg = preprocessing(cfg)

% combine EEG and EMG in one dataset, both are referenced differently
data = appenddata([], data_eeg, data_emg);

cfg = ... frequency analysis stuff
cfg.ouptu = 'powandcsd'
cfg.channelcmb = {'all', 'all'}
freq = freqanalysis(cfg, data);

cfg = ... source specific settings
cfg.refchan = one of your bipolar EMG channels
cfg.channel = all your EEG channels
source = sourceanalysis(cfg, freq)

best regards,
Robert


From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:46:59 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:46:59 +0200
Subject: import bva time-freq data
In-Reply-To: <LISTSERV%200607251739327460.3C55@NIC.SURFNET.NL>
Message-ID: <WED.26.JUL.2006.094659.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Stephan,

On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
> I try to import brainvision analyzer dat files of wavelet time-
> frequency
> decompositions into fieldtrip. I exported an average file, so it
> consists
> only of a single segment. However I have some problems with it I am
> not sure
...
> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
> \WT_all_ft_test.vmrk';
> cfg.trialdef.stim  = 'New Segment';
> %cfg.trialdef.segment  = 'yes';
> cfg.trialdef.timezero = 'yes';
> % cfg.trialdef.eventtype  = 'Time 0';

In general I would recommend to use the trialfun_general instead of
the trialfun_brainvision (but the latter should also work). The
trialfun_general was written later and will work for any dataformat
that is supported by read_fcdc_event, the trialfun_brainvision is
only there to support old scripts (i.e. scripts that predate the
read_fcdc_event function).

If not specified manually as trialfun, the trialfun_general or
trialfun_brainvision is selected automatically based on the cfg
settings. In your case you could do
    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
    cfg.trialdef.prestim    = number, latency in seconds (optional)
    cfg.trialdef.poststim   = number, latency in seconds (optional)
See teh help of DEFINETRIAL.

But in your case this is not the problem. The problem lies in
> cfg.trialdef.prestim    = 0.500;
> cfg.trialdef.poststim   = 1.600;

If I do
   event = read_fcdc_event('WT_all_ft_test.vmrk')
then I see that

 >> event(1)
ans =
         type: 'New Segment'
        value: '1'
       sample: 1
     duration: 0
       offset: []

 >> event(2)
ans =
         type: 'Time 0'
        value: '251'
       sample: 1
     duration: 0
       offset: []

The 'New Segment' event is at sample 1, hence you cannot select a
500ms pre-stimulus window before it (you cannot read before the
beginning of the data). The alternative 'Time 0' event is also at
sample 1. It specifies a value of 251, which surprises me. I would
expect that event to be present at sample 251.

Reading the marker file with a text editor, I see (abbreviated) the
explanation
    Mk<number>=<Type>,<Description>,<Position>,<Size>,
<Channelnumber>,<Date>
and the markers
    Mk1=New Segment,,1,1,0,00000000000000000000
    Mk2=Time 0,,251,1,0
So Mk2 should indeed be at sample 251.

With the matlab debugger in the private/read_event.m file (around
line 164) I notice that the problem lies in the missing value of the
second field, i.e.
    Mk2=Time 0,,251,1,0
is interpreted as
    Mk2=Time 0,251,1,0  (one comma less)
All fields are shifted by one, resulting in the sample number being
interpreted as the description (i.e. value).

I have fixed it in read_event, by replacing line 151 from
   tok = tokenize(line, '=');
into
   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
seperator
and by replacing line 157 from
   tok = tokenize(tok{2}, ',');
into
   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
the seperator
i.e. both calls to the tokenize function now use the third argument.
You can apply the same change to your copy, but please check that you
have an up to date version of tokenize that accepts 3 input arguments
(tokenize is in private as well). I will include the bug fix in the
upcoming nightly release of FieldTrip. Better upgrade tomorrow to the
latest FT version from the FTP server.

I think that sofar it was not noticed since people here at the
Donders typically use triggers as markers in their Brainvision data,
and in case of triggers the second field would not be empty.

If I now do
   cfg = [];
   cfg.dataset = 'WT_all_ft_test.vhdr'
   cfg.trialdef.eventtype = 'Time 0'
   cfg.trialdef.prestim  =  0.5000
   cfg.trialdef.poststim =  1
   cfg = definetrial(cfg)
I get
   evaluating trialfunction 'trialfun_general'
   found 2 events
   created 1 trials
   cfg =
        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: []
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
which is correct. The trial (cfg.trl) runs from sample 1 to sample
750 in the data file, and the first sample of that trial corresponds
with time cfg.trl(1,3)/fsample=-0.500 seconds.

Thanks for reporting the bug,
Robert


From stephan.bickel at ANATOM.UNIZH.CH  Wed Jul 26 17:28:54 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel, Anatomisches Inst.)
Date: Wed, 26 Jul 2006 11:28:54 -0400
Subject: import bva time-freq data
In-Reply-To: <30468A0B-0B5C-41EF-9FB3-84FD50036A22@fcdonders.ru.nl>
Message-ID: <WED.26.JUL.2006.112854.0400.FIELDTRIP@NIC.SURFNET.NL>

Hi Robert,

thank you very much for your response and for fixing the bug. Reading in the
trial information works fine now. Unfortunately I encounter another problem
when I want to read in the data with preprocessing. I wonder if it is yet
possible to read in time-frequency data from bva-dat files?

Sorry for bothering you again and thank you very much in advance.

Stephan


So when I do:

cfg = [];
cfg.dataset = 'WT_all_ft_test.vhdr'
cfg.trialdef.eventtype = 'Time 0'
cfg.trialdef.prestim = 0.5000
cfg.trialdef.poststim = 1
cfg = definetrial(cfg)
cfg.datafile   = 'WT_all_ft_test.dat';
cfg.channel = 'all';
[raw] = preprocessing(cfg)


I get:

cfg =

        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: 'WT_all_ft_test.dat'
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
        channel: 'all'

retaining exist trial definition
retaining exist event information
found 2 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
     dat(line,:) = str2num(str);

Error in ==> fieldtrip-20060725\private\read_data at 239
     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Error in ==> read_fcdc_data at 49
[dat] = read_data(varargin{:});

Error in ==> preprocessing at 366
   dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);



On Wed, 26 Jul 2006 09:46:59 +0200
  Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> wrote:
> Hi Stephan,
>
> On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
>> I try to import brainvision analyzer dat files of wavelet time-
>> frequency
>> decompositions into fieldtrip. I exported an average file, so it
>> consists
>> only of a single segment. However I have some problems with it I am
>> not sure
> ...
>> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
>> \WT_all_ft_test.vmrk';
>> cfg.trialdef.stim  = 'New Segment';
>> %cfg.trialdef.segment  = 'yes';
>> cfg.trialdef.timezero = 'yes';
>> % cfg.trialdef.eventtype  = 'Time 0';
>
> In general I would recommend to use the trialfun_general instead of  the
>trialfun_brainvision (but the latter should also work). The
> trialfun_general was written later and will work for any dataformat  that
>is supported by read_fcdc_event, the trialfun_brainvision is  only there to
>support old scripts (i.e. scripts that predate the  read_fcdc_event
>function).
>
> If not specified manually as trialfun, the trialfun_general or
> trialfun_brainvision is selected automatically based on the cfg  settings.
>In your case you could do
>    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
>    cfg.trialdef.prestim    = number, latency in seconds (optional)
>    cfg.trialdef.poststim   = number, latency in seconds (optional)
> See teh help of DEFINETRIAL.
>
> But in your case this is not the problem. The problem lies in
>> cfg.trialdef.prestim    = 0.500;
>> cfg.trialdef.poststim   = 1.600;
>
> If I do
>   event = read_fcdc_event('WT_all_ft_test.vmrk')
> then I see that
>
> >> event(1)
> ans =
>         type: 'New Segment'
>        value: '1'
>       sample: 1
>     duration: 0
>       offset: []
>
> >> event(2)
> ans =
>         type: 'Time 0'
>        value: '251'
>       sample: 1
>     duration: 0
>       offset: []
>
> The 'New Segment' event is at sample 1, hence you cannot select a  500ms
>pre-stimulus window before it (you cannot read before the  beginning of the
>data). The alternative 'Time 0' event is also at  sample 1. It specifies a
>value of 251, which surprises me. I would  expect that event to be present
>at sample 251.
>
> Reading the marker file with a text editor, I see (abbreviated) the
> explanation
>    Mk<number>=<Type>,<Description>,<Position>,<Size>,
> <Channelnumber>,<Date>
> and the markers
>    Mk1=New Segment,,1,1,0,00000000000000000000
>    Mk2=Time 0,,251,1,0
> So Mk2 should indeed be at sample 251.
>
> With the matlab debugger in the private/read_event.m file (around  line
>164) I notice that the problem lies in the missing value of the  second
>field, i.e.
>    Mk2=Time 0,,251,1,0
> is interpreted as
>    Mk2=Time 0,251,1,0  (one comma less)
> All fields are shifted by one, resulting in the sample number being
> interpreted as the description (i.e. value).
>
> I have fixed it in read_event, by replacing line 151 from
>   tok = tokenize(line, '=');
> into
>   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
> seperator
> and by replacing line 157 from
>   tok = tokenize(tok{2}, ',');
> into
>   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
> the seperator
> i.e. both calls to the tokenize function now use the third argument.  You
>can apply the same change to your copy, but please check that you  have an
>up to date version of tokenize that accepts 3 input arguments  (tokenize is
>in private as well). I will include the bug fix in the  upcoming nightly
>release of FieldTrip. Better upgrade tomorrow to the  latest FT version
>from the FTP server.
>
> I think that sofar it was not noticed since people here at the  Donders
>typically use triggers as markers in their Brainvision data,  and in case
>of triggers the second field would not be empty.
>
> If I now do
>   cfg = [];
>   cfg.dataset = 'WT_all_ft_test.vhdr'
>   cfg.trialdef.eventtype = 'Time 0'
>   cfg.trialdef.prestim  =  0.5000
>   cfg.trialdef.poststim =  1
>   cfg = definetrial(cfg)
> I get
>   evaluating trialfunction 'trialfun_general'
>   found 2 events
>   created 1 trials
>   cfg =
>        dataset: 'WT_all_ft_test.vhdr'
>       trialdef: [1x1 struct]
>       datafile: []
>     headerfile: 'WT_all_ft_test.vhdr'
>       trialfun: 'trialfun_general'
>          event: [1x2 struct]
>            trl: [1 750 -250]
>        version: [1x1 struct]
> which is correct. The trial (cfg.trl) runs from sample 1 to sample  750 in
>the data file, and the first sample of that trial corresponds  with time
>cfg.trl(1,3)/fsample=-0.500 seconds.
>
> Thanks for reporting the bug,
> Robert


From r.oostenveld at FCDONDERS.RU.NL  Thu Jul 27 09:26:26 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Thu, 27 Jul 2006 09:26:26 +0200
Subject: import bva time-freq data
In-Reply-To: <web-4062129@idmailbe2.unizh.ch>
Message-ID: <THU.27.JUL.2006.092626.0200.FIELDTRIP@NIC.SURFNET.NL>

On 26 Jul 2006, at 17:28, Stephan Bickel, Anatomisches Inst. wrote:
> thank you very much for your response and for fixing the bug.
> Reading in the trial information works fine now. Unfortunately I
> encounter another problem when I want to read in the data with
> preprocessing. I wonder if it is yet possible to read in time-
> frequency data from bva-dat files?

...

> Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
>     dat(line,:) = str2num(str);
>
> Error in ==> fieldtrip-20060725\private\read_data at 239
>     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Hi Stephan,

I have no idea how time-frequency data is stored in the BVA *.dat
files. Even if the low-level reading function were to support it, the
PREPROCESSING function at least would not be able to deal with it,
since the data structure returned by preprocessing does not allow for
3D data (channelXtimeXfrequency). That type of data would have to be
stored in a structure that is compatible with the output of
FREQANALYSIS. The preprocessing function is made for raw continuous
(or trial-based) data, and it happens to work as well for averaged
ERP data, although it is not specifically made for that.

For other external software (BESA, EEGLAB, LORETA) I have created
xxx2fieldtrip functions, i.e. functions that import already processed
data and format it into the fieldtrip structure that is most
compatible with it. In this case I can imagine a
BRAINVISION2FIELDTRIP function that would read your dat file and
return a "freq" structure. Besides making a brainvision2fieldtrip
function, it requires making either the read_brainvision_eeg smarter
(or probably making a seperate read_brainvision_tfr function) and
improving the detection of teh content of the BVA *.dat file. Please
have a look in besa2fieldtrip to see how I approach it there.

best regards,
Robert


From jciveira at UNAV.ES  Thu Jul 27 18:54:27 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Thu, 27 Jul 2006 18:54:27 +0200
Subject: Source anaylsis
In-Reply-To: <9A0AA0F8-A8AD-4958-B224-4E3547238053@FCDONDERS.RU.NL>
Message-ID: <THU.27.JUL.2006.185427.0200.FIELDTRIP@NIC.SURFNET.NL>

   Dear Robert:

   Thanks for your answer but my problems are still there. The main
problem is that i can not use DICS with a reference channel on EEG
data, no matter if  the reference is an EEG channel or EMG or some
other channel.

   I have taken my measures with 85 electrodes, some of them are
EEG1010, some EOG and some other that I have placed myself.

   I do the freq analysis and then the source analysis. Then I can
make another source analysis with refdip but there is always errors
when I use refchan. I do not think is a problem of the reference ,
cos always the program says that the  cross-spectral-density with
the reference channel is not complete .

   I have done the freq analysis making cfg.channelcmb combinations of
all channels with only one, so i make sure that the
cross-spectral-density is complete,but when I make source analysis
with that channel as a reference is the same.

   Searching in the code I have come up with the fact that the error
is always the condition in line 195 of the function
prepare_freq_matrices (version 2006/03/08). My idea is that there is
a confussion because if we have a set of N channels and we make  
combinations with channelcombination, the result is N-1 combinations  
because we do not include the combination of the reference with itself.

I hope i explain myself clear enough, anyway I am trying to fix the  
problem by myself but any help or suggestion will be great

Juan


Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
>> I have the same problem and I do not know how to fix it, even   
>> though I tried your solution with cfg.channelcmb.
>>
>> My case is : i have a set of 82 electrodes, I make freqanalysis   
>> with cfg.channelcmb = channelcombination({'all'
'all'},data.label);
>>  thus I get
>>
>> freq =
>>
>>        label: {82x1 cell}
>>       dimord: 'chan_freq'
>>         freq: [14 14.4000 14.8000 15.2000 15.6000 16
16.4000
>> 16.8000 17.2000 17.6000 18]
>>    powspctrm: [82x11 double]
>>     labelcmb: {3321x2 cell}
>>    crsspctrm: [3321x11 double]
>>          cfg: [1x1 struct]
>>
>> When I make sourceanalysis with a dipole as a reference and I
have
>> no problems but when I want to use a channel as a reference I   
>> always have the same error:
>> The cross-spectral-density matrix is not complete.
>>
>> Can you help me?
>>
>> Best regards
>> Juan
>
> Dear Juan
>
> Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
> expect the latter, since computing coherence between one particular
EEG
> electrode and sources in the brain with DICS will not lead to
> meaningfull results. The selected EEG electrode will pick up
> volume-conducted activity from all sources in the brain (as opposed
to
> an EMG electrode) and would be referenced to the same reference
> electrode as all other electrodes. So there will be a huge amount
of
> trivial coherence, whose spatial distribution in the brain will
> probably more reflect the volume-conduction than something
> physiologically interpretable.
>
> Although I have not used DICS with a reference channel on EEG data,
I
> would expect the code to work if you follow a procedure like this:
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EEG electrodes
> cfg.implicitref = name of implicit reference electrode used for the
EEG
> channels
> cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
> data_eeg = preprocessing(cfg)
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EMG electrodes, assuming that they are
> bipolar channels
> % cfg.reref may be used if you have two unipolar channels on the
muscle
> of interest
> cfg.rectify = 'yes'   % this makes the spiky nature of the EMG
more
> visible, and improves the coherence estimate
> cfg.baseline = 'yes'  % remove baseline prior to rectification,
not
> neccessary but still preferable
> data_emg = preprocessing(cfg)
>
> % combine EEG and EMG in one dataset, both are referenced
differently
> data = appenddata([], data_eeg, data_emg);
>
> cfg = ... frequency analysis stuff
> cfg.ouptu = 'powandcsd'
> cfg.channelcmb = {'all', 'all'}
> freq = freqanalysis(cfg, data);
>
> cfg = ... source specific settings
> cfg.refchan = one of your bipolar EMG channels
> cfg.channel = all your EEG channels
> source = sourceanalysis(cfg, freq)
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 09:49:55 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 09:49:55 +0200
Subject: Source anaylsis
In-Reply-To: <20060727185427.3r2gdy6zkk0o4og8@webmail.unav.es>
Message-ID: <FRI.28.JUL.2006.094955.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Juan

>   I do the freq analysis and then the source analysis. Then I can
> make another source analysis with refdip but there is always errors
> when I use refchan. I do not think is a problem of the reference ,
> cos always the program says that the  cross-spectral-density with
> the reference channel is not complete .
>
>   I have done the freq analysis making cfg.channelcmb combinations of
> all channels with only one, so i make sure that the
> cross-spectral-density is complete,but when I make source analysis
> with that channel as a reference is the same.
>
>   Searching in the code I have come up with the fact that the error
> is always the condition in line 195 of the function
> prepare_freq_matrices (version 2006/03/08). My idea is that there is
> a confussion because if we have a set of N channels and we make
> combinations with channelcombination, the result is N-1
> combinations because we do not include the combination of the
> reference with itself.

Say for simplicity that you want EMG as refchan. In freqanalysis,
cfg.channelcmb should be all with all. You need the CSD between all
EEG channels (all EEG with all EEG), and of all EEG channels with the
EMG. In prepare_freq_matrices the CSD matrix (Cf) of EEG with EEG is
made (Neeg X Neeg), the CSD vector (Cr) of all EEG channels with the
reference EMG channel (Neeg X 1), and the power of the reference
channel (Pr). Together they could also be shaped in a (Neeg+1) X (Neeg
+1) matrix in principle, but that is not done for DICS.

I have tried it out myself, and cannot reproduce your problem. Please
find my test script below. It uses some simulated EEG data. The
beginning of the script is copied from http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:compute_forward_simulated_data_and_apply_a_di
pole_fit

The initial part generates raw EEG data, then I add a EMG channel
(with noise, but it is only to demonstrate that fieldtrip does not
give an error), then I do freqanalysis and sourceanalysis. Please try
out and compare my cfgs with yours, since I suspect your
configuration to be the problem.

best regards,
Robert


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% create a set of electrodes, randomly placed on the sphere
elec = [];
elec.pnt = randn(128,3);
dum = sqrt(sum(elec.pnt.^2,2));
elec.pnt = elec.pnt ./ [dum dum dum];  % scale them to a unit sphere
for i=1:128
    elec.label{i} = sprintf('%03d', i);
end

% create a concentric 3-sphere volume conductor, the radius is the
same as for the electrodes
vol = [];
vol.r = [0.88 0.92 1.00]; % radii of spheres
vol.c = [1 1/80 1];       % conductivity
vol.o = [0 0 0];          % center of sphere

% create a dipole simulation with one dipole and a 10Hz sine wave
cfg      = [];
cfg.vol  = vol;             % see above
cfg.elec = elec;            % see above
cfg.dip.pos = [0 0.5 0.3];
cfg.dip.mom = [1 0 0]';     % note, it should be transposed
cfg.dip.frequency = 10;
cfg.ntrials = 10;
cfg.triallength = 1;        % seconds
cfg.fsample = 250;          % Hz
raw1 = dipolesimulation(cfg);

using headmodel specified in the configuration
using electrodes specified in the configuration
selected 128 electrodes
computing simulated data
RMS value of simulated data is 0.0746553

%%%%%%%%%%%%%%%%%%%%%%%%
%%%%% Here it starts deviating from the demo script on the wiki %%%%
%%%%%%%%%%%%%%%%%%%%%%%%

% add a fake EMG channel, containing noise
for i=1:10
   raw1.trial{i}(end+1,:) = randn(1,250);
end
raw1.label{end+1} = 'EMG'

raw1 =

       trial: {1x10 cell}
        time: {1x10 cell}
        elec: [1x1 struct]
     fsample: 250
       label: {1x129 cell}
         cfg: [1x1 struct]

cfg = []
cfg.method = 'mtmfft';
cfg.taper = 'hanning';
cfg.foilim = [2 20];
cfg.output = 'powandcsd';
cfg.channelcmb = {'all', 'all'}

cfg =

         method: 'mtmfft'
          taper: 'hanning'
         foilim: [2 20]
         output: 'powandcsd'
     channelcmb: {'all'  'all'}

freq1 = freqanalysis(cfg, raw1)
WARNING: using only one taper for specified smoothing
processing trial 1, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 2, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 3, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 4, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 5, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 6, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 7, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 8, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 9, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 10, nfft: 250 samples, taper length: 250 samples, 1
tapers

freq1 =
          label: {1x129 cell}
        dimord: 'chan_freq'
          freq: [2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20]
     powspctrm: [129x19 double]
      labelcmb: {8256x2 cell}
     crsspctrm: [8256x19 double]
          elec: [1x1 struct]
           cfg: [1x1 struct]

cfg = [];
cfg.vol = vol;
cfg.freq = 10;
cfg.frequency = 10;
cfg.method = 'dics';
cfg.refchan = 'EMG';
cfg.grid.resolution = 0.3

cfg =
           vol: [1x1 struct]
          freq: 10
     frequency: 10
        method: 'dics'
       refchan: 'EMG'
          grid: [1x1 struct]

source1 = sourceanalysis(cfg, freq1)
using headmodel specified in the configuration
using electrodes specified in the data
selected 128 electrodes
106 dipoles inside, 237 dipoles outside brain
making tight grid
106 dipoles inside, 110 dipoles outside brain
scanning repetition 1
scanning grid
Warning: cross-spectral density matrix is rank deficient
 > In <a href="error:/Users/roberto/matlab/forwinv/beamformer.m,
270,1">beamformer at 270</a>
   In <a href="error:/Users/roberto/matlab/fieldtrip/sourceanalysis.m,
806,1">sourceanalysis at 806</a>
total time in sourceanalysis 75.4 seconds

source1 =
         xgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         ygrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         zgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
           dim: [6 6 6]
           vol: [1x1 struct]
     frequency: 10
           pos: [216x3 double]
        inside: [1x106 double]
       outside: [1x110 double]
        method: 'average'
           avg: [1x1 struct]
           cfg: [1x1 struct]

diary off


From floris.delange at FCDONDERS.RU.NL  Fri Jul 28 13:01:00 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Fri, 28 Jul 2006 13:01:00 +0200
Subject: sourcestatistics: what to compare?
Message-ID: <FRI.28.JUL.2006.130100.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Fieldtrippers,

I have a question about what conditions/comparisons make most sense to
statistically compare with sourcestatistics.
I have 2 conditions, A and B, and each condtion has its own pre-stimulus
baseline: basA and basB.

Now, first I'd like to see which regions are showing stronger beta-band
power during A than the baseline. So I want to compare A to basA.
There's 2 ways of going about it:
1) compare (the grand average of) A.avg.pow with basA.avg.pow
2) compare (the grand average of) the neural activity index (NAI) of A
with the baseline

Any ideas which would be best? Or is this just an empirical issue?
(i.e., take the best)

To compare A and B, one has even more options:
1) compare A.avg.pow with B.avg.pow
2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
3) compare NAI(A) with NAI(B)
4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)

Is there's any principled reason for a priori picking any of the above
choices?

Your input is greatly appreciated,
Kind regards,

Floris


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 16:17:24 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 16:17:24 +0200
Subject: sourcestatistics: what to compare?
In-Reply-To: <44C9EE6C.8030003@fcdonders.ru.nl>
Message-ID: <FRI.28.JUL.2006.161724.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris

On 28 Jul 2006, at 13:01, Floris de Lange wrote:
> I have a question about what conditions/comparisons make most sense
> to statistically compare with sourcestatistics.
> I have 2 conditions, A and B, and each condtion has its own pre-
> stimulus baseline: basA and basB.
>
> Now, first I'd like to see which regions are showing stronger beta-
> band power during A than the baseline. So I want to compare A to
> basA. There's 2 ways of going about it:
> 1) compare (the grand average of) A.avg.pow with basA.avg.pow
> 2) compare (the grand average of) the neural activity index (NAI)
> of A with the baseline
>
> Any ideas which would be best? Or is this just an empirical issue?
> (i.e., take the best)

Both are valid, you could take the best. There are some exceptions/
predictions though that I can share with you: the NAI is power
divided by estimated (projected) noise. It is inherently difficult to
estimate the noise level (the default is the smallest singular value
of the CSD matrix, but you can manipulate it with cfg.lambda in
sourceanalysis). If you have different numbers of trials in active
v.s. baseline, the noise estimate probably will be different. So then
you would have NAIa = Pa/Na and NAIb = Pb/Nb, if Pa==Pb and Na~=Nb,
these will be trivially different, which is not what you want
(remember Pa==Pb was assumed).

You can compare the projected noise in both conditions, the source-
noise-estimate is uniformely scaled (i.e. on all voxels the same)
with the estimated noise from the CSD. There is no reason to assume
that the noise is different, so if the noise is different, you would
have to explain that.

> To compare A and B, one has even more options:
> 1) compare A.avg.pow with B.avg.pow
> 2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
> 3) compare NAI(A) with NAI(B)
> 4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)
>
> Is there's any principled reason for a priori picking any of the
> above choices?

I presume that by "compare" you mean "look at the the difference".

In general you should only use the NAI for plotting data if you only
have one condition (and no baseline). The noise estimate that goes
into the NAI is too poor to be of real value. Although sometines it
looks nice in figures, the NAI tends not to be very robust (it
contains global flucuations that mess up the statistic, although the
figure still looks nice).

The baseline is only interesting if you expect something differently
to happen in the two baselines. If that is the case, you better try
to get another baseline in your experimental design. If you would
combine actA, basA, actB, basB into a single number, and that number
would be significant, you still would not be able to tell whether the
difference is betwen the active conditions, theis baselines, or some
weird combination of the 4 "conditions" that you combine.

Typically there is no reason to assume taht the baselines are
different. So
   (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
= (A.avg.pow/basC.avg.pow) with (B.avg.pow/basC.avg.pow), where C
means common
= (A.avg.pow-B.avg.pow)/basC.avg.pow.
I.e. you are comparing A and B directly, with a spatially normalizing
for the different baseline-estimated noise at different source
locations.

Whether you want to spatially normalize (divide by some baseline
estimated-source-power) depends on the leadfields (it is neccessary
if cfg.normalize=no, if you want to do multiple comparison
correction, and if you want to look at an additive effect). If you
want to look at a multiplicative effect, you would look at the ratio
between two conditions instead of the difference. The ratio of the
conditions is best implemented with a log transform, after which you
again can look at the difference between the log-transformed power
estimates in both conditions (log(A/B) = log(A) - log(B)). In the
ratio or log-ratio between the powers in both conditions you have
also corrected for the spatial (depth) noise-bias.

good luck with the stats,
Robert

PS as I am on holiday the next 2 weeks, you could ask Ingrid about
more details on this. She is more or less at the same stage in her
analysis (about to do group sourcestatistics), but I suspect that she
already has more "feel" for these issues. Or ask JM.




From d.vanbarneveld at STUDENT.SCIENCE.RU.NL  Thu Jul  6 16:58:40 2006
From: d.vanbarneveld at STUDENT.SCIENCE.RU.NL (Denise van Barneveld)
Date: Thu, 6 Jul 2006 16:58:40 +0200
Subject: problem with text in topoplot and triplot
Message-ID: <THU.6.JUL.2006.165840.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear all,

since I started using the new version of Matlab (R2006a) on a mac, I
have this problem with text in the figures which I generate with
topoplot.m and triplot.m. The upper half of the text lines (title,
colorbar, electrodes, etc) is not visible (see attached figure).
Does anyone have this problem too? Especially people who are using
this new version on a pc. And does anyone perhaps has a solution for
this problem? I am currently looking in the code, if I can find where
it goes wrong, but I wonder whether it is a problem only with the new
version, or only at a mac, or only at my computer.

Best regards,
Denise van Barneveld

-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: application/applefile
Size: 48798 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment-0001.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: image/jpeg
Size: 85086 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment-0001.jpg>

From muthuraman10 at HOTMAIL.COM  Fri Jul  7 12:04:26 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Fri, 7 Jul 2006 10:04:26 +0000
Subject: query on confidence limit, DICS
In-Reply-To: <37FAA4DD-5DD8-4F2F-ABA6-D9F47F6AB95E@fcdonders.ru.nl>
Message-ID: <FRI.7.JUL.2006.100426.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

As i am able to sucessfully use the coherence analysis using freq_mtmfft and
frq_mtmwelch to calculate the coherence
i would like to know is there a way to calculate the confidence limit for
the coherence calculated from the freqanalysis, is there a standard code

Next, i would like to follow with the source analysis 'DICS' is there a
tutorial available

thanking you

with regards
Muthuraman


>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on CTF software!!!
>Date: Fri, 23 Jun 2006 15:23:22 +0200
>
>dear Muthuraman
>
>On 23 Jun 2006, at 12:54, Muthuraman Muthuraman wrote:
>>For the Beamformer source analysis, do we need the CTF software
>>because when i am trying out the tutorial in beamforming
>
>You do not need the ctf software to follow the tutorial. The tutorial  is
>based however on a ctf MEG dataset.
>
>>CTF_READ_RES4 [v  1.12]
>>....done (  0.67 sec)
>>
>>i get the error
>>
>>??? Error using ==> read_fcdc_header
>>could not read CTF res4 header file
>>
>>Error in ==> preprocessing at 263
>>hdr = read_fcdc_header(cfg.headerfile);
>
>Based on your error I suspect that matlab/fieldtrip cannot find the
>dataset that you are trying to analyze, or that you matlab path is
>incorrect. The function "CTF_READ_RES4" is not supposed to be called  (it
>is part of the NIH ctf import routines, e.g. used in EEGLAB),  since the
>default private/read_ctf_res4.m function can deal with the  tutorial
>dataset just fine. Please look into the code of  read_fcdc_header (using
>the matlab debugger) to figure out what your  exact problem is.
>
>best
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul  7 12:36:02 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 7 Jul 2006 12:36:02 +0200
Subject: query on confidence limit, DICS
In-Reply-To: <BAY121-F19F77F86077022EC6849B3CE740@phx.gbl>
Message-ID: <FRI.7.JUL.2006.123602.0200.FIELDTRIP@NIC.SURFNET.NL>

On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
> As i am able to sucessfully use the coherence analysis using
> freq_mtmfft and frq_mtmwelch to calculate the coherence
> i would like to know is there a way to calculate the confidence
> limit for the coherence calculated from the freqanalysis, is there
> a standard code

You can compute a jackknife estimate of the variance in the coherence
using the freqdescriptives function. That requires keeptrials=yes in
freqanalysis.

> Next, i would like to follow with the source analysis 'DICS' is
> there a tutorial available

No there is no tutorial available specifically for that. You can look
at the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
options are cfg.method=dics, cfg.refchan and/or cfg.refdip.

Robert


From j.poort at NIN.KNAW.NL  Tue Jul 11 11:43:49 2006
From: j.poort at NIN.KNAW.NL (Jasper Poort)
Date: Tue, 11 Jul 2006 11:43:49 +0200
Subject: Confidence intervals for evoked potentials
Message-ID: <TUE.11.JUL.2006.114349.0200.FIELDTRIP@NIC.SURFNET.NL>

 
Hi all, I have some questions regarding confidence intervals for timelocked data in FieldTrip. I noted that a field .var is returned by timelockanalysis

line 510 timelockanalysis
% compute the variance
tmp1 = repmat(dofvec(:)', [nchan 1]);
tmp2 = repmat(dofvec(:)', [nchan 1])-1;
var = (ss - (s.^2)./tmp1) ./ tmp2;

Assuming that the sampling distribution is normal I can presumably use this field for computation of confidence intervals of the .avg field? 
Is nonparametric estimation of variance using a bootstrap or jackknife also implemented somewhere? In general I'm just curious to know how others compute confidence intervals for evoked potentials in FieldTrip, for example to get an idea at what point in time two conditions start to differ, 

best regards, Jasper
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060711/806a2020/attachment-0001.html>

From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 15:50:09 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 15:50:09 +0200
Subject: Confidence intervals for evoked potentials
In-Reply-To: <017d01c6a4ce$835853e0$910b57c0@ABSYS>
Message-ID: <TUE.11.JUL.2006.155009.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Jasper,

On 11 Jul 2006, at 11:43, Jasper Poort wrote:
> Hi all, I have some questions regarding confidence intervals for
> timelocked data in FieldTrip. I noted that a field .var is returned
> by timelockanalysis
>
> line 510 timelockanalysis
> % compute the variance
> tmp1 = repmat(dofvec(:)', [nchan 1]);
> tmp2 = repmat(dofvec(:)', [nchan 1])-1;
> var = (ss - (s.^2)./tmp1) ./ tmp2;
>
> Assuming that the sampling distribution is normal I can presumably
> use this field for computation of confidence intervals of the .avg
> field?

Yes, you can use it to compute the channel-timepoint specific
standard-error of mean (SEM).

> Is nonparametric estimation of variance using a bootstrap or
> jackknife also implemented somewhere? In general I'm just curious
> to know how others compute confidence intervals for evoked
> potentials in FieldTrip, for example to get an idea at what point
> in time two conditions start to differ,

Bootstrap and jackknife estimates of variance are only different from
teh normal variance estimate if the measure involves a nonlinear or
biased estimate (i.e. coherence). The jackknife estimate of variance
of the ERP is exactly the same as the normal estimate of variance,
since the ERP is a plain (linear) average. The same holds for
bootstrap estimates.

So that means that for ERPs the jackknife and bootstrap only would be
a computationally expensive methods, resulting in the same variance
as usual. Therefore they are not implemented.

best regards,
Robert


From muthuraman10 at HOTMAIL.COM  Tue Jul 11 16:59:58 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 11 Jul 2006 14:59:58 +0000
Subject: DICS
In-Reply-To: <E2FC4366-C6DD-476C-B902-6EBF864D6B48@fcdonders.ru.nl>
Message-ID: <TUE.11.JUL.2006.145958.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thanks for the suggestion to use the jackknife variance estimate it works
thanks a lot

To use the DICS method

We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG channels
recording
and also we have the 64 channel measurement in FRMI

the electrode locations we have it in a file with 3 columns x,y,z for each
electrode
using the polhemus sensors but our own software

In the Source analysis
Cfg.method='DICS';
but Cfg.xgrid, ygrid and zgrid
the grid values are the sensor location values

or it should be the
cfg.grid.pos-where i can give in the 3 values [x y z] for each electrode

I am a starter in the source analysis in combining EEG-FMRI
please let me know

with regards
muthuraman.






>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on confidence limit, DICS
>Date: Fri, 7 Jul 2006 12:36:02 +0200
>
>On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
>>As i am able to sucessfully use the coherence analysis using  freq_mtmfft
>>and frq_mtmwelch to calculate the coherence
>>i would like to know is there a way to calculate the confidence  limit for
>>the coherence calculated from the freqanalysis, is there  a standard code
>
>You can compute a jackknife estimate of the variance in the coherence
>using the freqdescriptives function. That requires keeptrials=yes in
>freqanalysis.
>
>>Next, i would like to follow with the source analysis 'DICS' is  there a
>>tutorial available
>
>No there is no tutorial available specifically for that. You can look  at
>the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
>doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
>options are cfg.method=dics, cfg.refchan and/or cfg.refdip.
>
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 17:51:28 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 17:51:28 +0200
Subject: DICS
In-Reply-To: <BAY121-F26D094D1C667F7EAEC07B0CE680@phx.gbl>
Message-ID: <TUE.11.JUL.2006.175128.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman

On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
> To use the DICS method
>
> We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG
> channels recording
> and also we have the 64 channel measurement in FRMI
>
> the electrode locations we have it in a file with 3 columns x,y,z
> for each electrode
> using the polhemus sensors but our own software
>
> In the Source analysis
> Cfg.method='DICS';
> but Cfg.xgrid, ygrid and zgrid
> the grid values are the sensor location values
>
> or it should be the
> cfg.grid.pos-where i can give in the 3 values [x y z] for each
> electrode

in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the
locations at which the _dipoles_ will be placed, it is not the
position of the electrodes. At each grid location a spatial filter is
constructed and the power (and coherence) is computed. If you do not
want to scan on a full 3-D grid, but on only a few dipole locations
of interest, you can use cfg.grid.pos to specify those dipole locations.

The electrode positions are specified in the cfg.elec structure, or
in the cfg.elecfile filename. You can find more details on the
definition of electrode positions at http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
_magnetometers_or_gradiometers_described

Furthermore, you can try using the READ_FCDC_ELEC function for
reading the electrode locations, but probably taht will not work
since youy are using your own software for digitization.
Alternatively, you can manually construct an "elec" electrode
definition.

best regards
Robert

PS please upgrade to a recent fieldtrip version, the sourceanalysis
options cfg.xgrid/ygrid/zgrid have been renamed. The old options
still will work, but in general it is preferable to work with a
recent copy of the toolbox.


From muthuraman10 at HOTMAIL.COM  Thu Jul 13 13:11:35 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Thu, 13 Jul 2006 11:11:35 +0000
Subject: DICS
In-Reply-To: <43B3E5A4-716B-48A0-8CE0-5F8810DC4A65@fcdonders.ru.nl>
Message-ID: <THU.13.JUL.2006.111135.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for the suggestions
I have dowloaded the latest version now and i am working with it
i have created the own elec file which is "elec1" and these are the steps i
follow
please go through it and reply me, why is the error

-First i calculate the coherence using the freqanalysis and freqdescriptives
which works perfectly
-Second i calculate the Grid using Prepare_leadfield

vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1);

which also works i get the output for the Grid

-Third I want use the source analysis for the DICS

cfg.method='dics';
cfg.grid=grid;
cfg.elecfile=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
??? Error using ==> fprintf
Function is not defined for 'struct' inputs.

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
  fprintf('reading electrodes from file %s\n', cfg.elecfile);

Error in ==> sourceanalysis at 574
[vol, sens, cfg] = prepare_vol_sens(cfg, data);




-I also tried grid using the freqmtmfft-which as the results for the
freqanalysis
vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1,freqmtmfft);

I get error

using headmodel specified in the configuration
using electrodes specified in the configuration
??? Error using ==> delaunayn
Not enough unique points to do tessellation.

Error in ==> delaunay at 49
    tri = delaunayn([x(:) y(:)]);

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));

Error in ==> prepare_leadfield at 179
[vol, sens, cfg] = prepare_vol_sens(cfg, data);


Thanking you

with regards
Muthuraman.




>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Tue, 11 Jul 2006 17:51:28 +0200
>
>Dear Muthuraman
>
>On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
>>To use the DICS method
>>
>>We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG  channels
>>recording
>>and also we have the 64 channel measurement in FRMI
>>
>>the electrode locations we have it in a file with 3 columns x,y,z  for
>>each electrode
>>using the polhemus sensors but our own software
>>
>>In the Source analysis
>>Cfg.method='DICS';
>>but Cfg.xgrid, ygrid and zgrid
>>the grid values are the sensor location values
>>
>>or it should be the
>>cfg.grid.pos-where i can give in the 3 values [x y z] for each  electrode
>
>in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the  locations at
>which the _dipoles_ will be placed, it is not the  position of the
>electrodes. At each grid location a spatial filter is  constructed and the
>power (and coherence) is computed. If you do not  want to scan on a full
>3-D grid, but on only a few dipole locations  of interest, you can use
>cfg.grid.pos to specify those dipole locations.
>
>The electrode positions are specified in the cfg.elec structure, or  in the
>cfg.elecfile filename. You can find more details on the  definition of
>electrode positions at http://www2.ru.nl/fcdonders/ fieldtrip/doku.php?
>id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
>_magnetometers_or_gradiometers_described
>
>Furthermore, you can try using the READ_FCDC_ELEC function for  reading the
>electrode locations, but probably taht will not work  since youy are using
>your own software for digitization.  Alternatively, you can manually
>construct an "elec" electrode  definition.
>
>best regards
>Robert
>
>PS please upgrade to a recent fieldtrip version, the sourceanalysis
>options cfg.xgrid/ygrid/zgrid have been renamed. The old options  still
>will work, but in general it is preferable to work with a  recent copy of
>the toolbox.


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 14 11:37:47 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 14 Jul 2006 11:37:47 +0200
Subject: DICS
In-Reply-To: <BAY121-F32B67944419449CAB9EEF6CE6E0@phx.gbl>
Message-ID: <FRI.14.JUL.2006.113747.0200.FIELDTRIP@NIC.SURFNET.NL>

On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:

> cfg.method='dics';
> cfg.grid=grid;
> cfg.elecfile=elec1;
> cfg.vol=vol;
> [source]=sourceanalysis(cfg,freqmtmfft);
> using headmodel specified in the configuration
> ??? Error using ==> fprintf
> Function is not defined for 'struct' inputs.
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>  fprintf('reading electrodes from file %s\n', cfg.elecfile);

You should specify either cfg.elecfile with a string (filename) or
cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)

> I get error
>
> using headmodel specified in the configuration
> using electrodes specified in the configuration
> ??? Error using ==> delaunayn
> Not enough unique points to do tessellation.
>
> Error in ==> delaunay at 49
>    tri = delaunayn([x(:) y(:)]);
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>
> Error in ==> prepare_leadfield at 179
> [vol, sens, cfg] = prepare_vol_sens(cfg, data);

That indicates that less than 3 electrodes were selected. Probably
there is a mismatch between the channel names in the data and the
names of the electrodes. FT selects the channels and electrodes based
on the corresponding labels, if they do not correspond then no
channels+electrodes are selected.

Robert


From muthuraman10 at HOTMAIL.COM  Mon Jul 17 16:12:14 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Mon, 17 Jul 2006 14:12:14 +0000
Subject: DICS
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <MON.17.JUL.2006.141214.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
but not when using the mtmwelch is this a problem in the mtmwelch
or i am doing something wrong

Please go through it

thanking you

with regards
muthuraman



>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Fri, 14 Jul 2006 11:37:47 +0200
>
>On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:
>
>>cfg.method='dics';
>>cfg.grid=grid;
>>cfg.elecfile=elec1;
>>cfg.vol=vol;
>>[source]=sourceanalysis(cfg,freqmtmfft);
>>using headmodel specified in the configuration
>>??? Error using ==> fprintf
>>Function is not defined for 'struct' inputs.
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>>  fprintf('reading electrodes from file %s\n', cfg.elecfile);
>
>You should specify either cfg.elecfile with a string (filename) or
>cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)
>
>>I get error
>>
>>using headmodel specified in the configuration
>>using electrodes specified in the configuration
>>??? Error using ==> delaunayn
>>Not enough unique points to do tessellation.
>>
>>Error in ==> delaunay at 49
>>    tri = delaunayn([x(:) y(:)]);
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>>
>>Error in ==> prepare_leadfield at 179
>>[vol, sens, cfg] = prepare_vol_sens(cfg, data);
>
>That indicates that less than 3 electrodes were selected. Probably  there
>is a mismatch between the channel names in the data and the  names of the
>electrodes. FT selects the channels and electrodes based  on the
>corresponding labels, if they do not correspond then no
>channels+electrodes are selected.
>
>Robert


From floris.delange at FCDONDERS.RU.NL  Tue Jul 25 10:24:18 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Tue, 25 Jul 2006 10:24:18 +0200
Subject: sourcestatistics question
Message-ID: <TUE.25.JUL.2006.102418.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear FieldTrippers,

I have a question concerning sourcestatistics.
I have calculated the source localization using DICS for 11 individual
subjects. Now I would like to test for the significance of the sources
using sourcestatistics. The examples I find in the tutorial are about
checking for significance within one subject by shuffling the individual
trials. However, what I would like to do is to do the statistics on the
2nd level (subject), and to test for the consistency of the source
reconstructions across subjects.
Is this possible with sourcestatistics? If so, how should I feed the
data to sourcestatistics? If I say something like stats =
sourcestatistics(cfg,sources{1:11}) where each source{} is a normalized
source (calculated by sourceanalysis, normalized with
sourceinterpolate), it complains (rightfully) that there's a reference
to the non-existing field 'statistic'.

Any ideas are greatly appreciated!

Kind regards,
Floris


From muthuraman10 at HOTMAIL.COM  Tue Jul 25 10:34:30 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 25 Jul 2006 08:34:30 +0000
Subject: Source anaylsis
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.083430.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
i get a error when using the mtmfft for source analysis
cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
using electrodes specified in the configuration
610 dipoles inside, 600 dipoles outside brain

??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
The cross-spectral-density matrix is not complete

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);


Please go through it
Because for the DICS i would like to use the freqmtmwelch
is this a problem with freqmtmwelch
or something wrong in my inputs

thanking you

with regards
muthuraman


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:18:48 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:18:48 +0200
Subject: DICS
In-Reply-To: <BAY121-F235CB8F83C35EABC33B41ECE620@phx.gbl>
Message-ID: <TUE.25.JUL.2006.121848.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:29:23 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:29:23 +0200
Subject: sourcestatistics question
In-Reply-To: <44C5D532.4020708@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.122923.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris,

On 25 Jul 2006, at 10:24, Floris de Lange wrote:
> I have a question concerning sourcestatistics.
> I have calculated the source localization using DICS for 11
> individual subjects. Now I would like to test for the significance
> of the sources using sourcestatistics. The examples I find in the
> tutorial are about checking for significance within one subject by
> shuffling the individual trials. However, what I would like to do
> is to do the statistics on the 2nd level (subject), and to test for
> the consistency of the source reconstructions across subjects.
> Is this possible with sourcestatistics?

Yes, that is possible.

> If so, how should I feed the data to sourcestatistics? If I say
> something like stats = sourcestatistics(cfg,sources{1:11}) where
> each source{} is a normalized source (calculated by sourceanalysis,
> normalized with sourceinterpolate), it complains (rightfully) that
> there's a reference to the non-existing field 'statistic'.


Assuming you want to use a randomization test to get a Monte-Carlo
estimate of the probability of your null-hypothesis
(cfg.method='montecarlo'), cfg.statistic should be a string
describing the statistic whose randomization distribution you want to
compute. See the help of statistic_montecarlo.

Furthermore, the hypothesis that you want to test is about two
conditions. That means that the two conditions both should be fed
into sourcestatistics, e.g. something like

for i=1:11
   read data from condition 1 for each subject
   cond1{i} = sourceanalysis(...)
   cond1{i} = sourceinterpolate(...)
   cond1{i} = volumenormalize(...)
   read data from condition 2 for each subject
   cond2{i} = sourceanalysis(...)
   cond2{i} = sourceinterpolate(...)
   cond2{i} = volumenormalize(...)
end

cfg = []
cfg.design = [1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2    %
condition number
               1 2 3 4 5 .....       1 2 3 4 5 ....       ]   %
subject number
cfg.method = 'montecarlo'
cfg.statistic = 'depsamplesT'
cfg.ivar = 1; % row with the independent variable (condition) in the
design
cfg.uvar = 2; % row with the unit of observation (subject) in the
design matrix
stat = sourcestatistics(cfg, cond1{:}, cond2{:});

It is also possible to do other tests, which are more in line with a
2nd level analysis (e.g. compute t-maps per subject and test over
subjects). It is better to discuss those face-to-face.

best regards,
Robert


From Jan.Schoffelen at FCDONDERS.RU.NL  Tue Jul 25 12:32:38 2006
From: Jan.Schoffelen at FCDONDERS.RU.NL (Jan Mathijs Schoffelen)
Date: Tue, 25 Jul 2006 12:32:38 +0200
Subject: DICS
In-Reply-To: <4442BD57-7F97-469F-990B-CD63E171E371@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.123238.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman,

One little addition to the freq.cumtapcnt-story.
freq.cumtapcnt should be a vector instead of a scalar:
the correct way to specify it would be: freq.cumtapcnt = ones(ntrl,1)
with ntrl the number of trials in your data.

Yours,

Jan-Mathijs

-----Original Message-----
From: FieldTrip discussion list [mailto:FIELDTRIP at NIC.SURFNET.NL] On Behalf
Of Robert Oostenveld
Sent: Tuesday, July 25, 2006 12:19 PM
To: FIELDTRIP at NIC.SURFNET.NL
Subject: Re: [FIELDTRIP] DICS

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:32:52 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:32:52 +0200
Subject: Source anaylsis
In-Reply-To: <BAY121-F324EA9761CB16729FB47E8CE5A0@phx.gbl>
Message-ID: <TUE.25.JUL.2006.123252.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:

> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
> The cross-spectral-density matrix is not complete

this indicates that you did not include all channel combinations
(cfg.channelcmb) in freqanalysis, and that you try to use a cross
spectral density matrix for DISC that is not complete (i.e. come
cross-spectra are missing). You should correct your configuration for
the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
CHANNELCOMBINATION function).

best regards,
Robert


From jciveira at UNAV.ES  Tue Jul 25 14:22:36 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Tue, 25 Jul 2006 14:22:36 +0200
Subject: Source anaylsis
In-Reply-To: <77E33A56-A0D4-43FC-94E0-91A3AC31197B@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.142236.0200.FIELDTRIP@NIC.SURFNET.NL>

I have the same problem and I do not know how to fix it, even though I  
tried your solution with cfg.channelcmb.

My case is : i have a set of 82 electrodes, I make freqanalysis with  
cfg.channelcmb = channelcombination({'all' 'all'},data.label);  thus I  
get

freq =

         label: {82x1 cell}
        dimord: 'chan_freq'
          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000 16.8000  
17.2000 17.6000 18]
     powspctrm: [82x11 double]
      labelcmb: {3321x2 cell}
     crsspctrm: [3321x11 double]
           cfg: [1x1 struct]

When I make sourceanalysis with a dipole as a reference and I have no  
problems but when I want to use a channel as a reference I always have  
the same error:
The cross-spectral-density matrix is not complete.

Can you help me?

Best regards
Juan

Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:
>
>> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
>> The cross-spectral-density matrix is not complete
>
> this indicates that you did not include all channel combinations
> (cfg.channelcmb) in freqanalysis, and that you try to use a cross
> spectral density matrix for DISC that is not complete (i.e. come
> cross-spectra are missing). You should correct your configuration for
> the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
> CHANNELCOMBINATION function).
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From stephan.bickel at ANATOM.UNIZH.CH  Tue Jul 25 17:39:32 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel)
Date: Tue, 25 Jul 2006 17:39:32 +0200
Subject: import bva time-freq data
Message-ID: <TUE.25.JUL.2006.173932.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi,

I try to import brainvision analyzer dat files of wavelet time-frequency
decompositions into fieldtrip. I exported an average file, so it consists
only of a single segment. However I have some problems with it I am not sure
if it is because I define the segment wrong or if it is not possible to import
this data into fieldtrip.
I pasted the code and the error message below and attached the marker- and
header file.

I would be really glad for any suggestions.

Thank you very much,

Stephan 



This is what I tried:

cfg = [];
cfg.headerfile =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vhdr';
cfg.datafile   =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.dat';
cfg.channel = 'all';

%Trialdefinition
cfg.trialdef.trgfile  =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vmrk';
cfg.trialdef.stim  = 'New Segment';
%cfg.trialdef.segment  = 'yes';
cfg.trialdef.timezero = 'yes';
% cfg.trialdef.eventtype  = 'Time 0';
cfg.trialdef.prestim    = 0.500;
cfg.trialdef.poststim   = 1.600;

[cfg] = definetrial(cfg)
[raw] = preprocessing(cfg)


and this is the error message:


evaluating trialfunction 'trialfun_brainvision'
0 stimulus markers converted into 0 trials
0 response markers converted into 0 trials
1 segment markers converted into 0 trials
1 "Time 0" markers converted into 1 trials
found 0 events
created 1 trials

cfg = 

    headerfile: [1x88 char]
      datafile: [1x87 char]
       channel: 'all'
      trialdef: [1x1 struct]
       dataset: []
      trialfun: 'trialfun_brainvision'
         event: []
           trl: [1 1051 -250]
       version: [1x1 struct]

retaining exist trial definition
retaining exist event information
found 0 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060330\private\read_brainvision_eeg at 65
    dat(line,:) = str2num(str);

Error in ==> read_fcdc_data at 419
  dat = read_brainvision_eeg(datafile, hdr, begsample, endsample);

Error in ==> preprocessing at 322
  dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);

-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vmrk
Type: application/octet-stream
Size: 741 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment-0002.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vhdr
Type: application/octet-stream
Size: 3459 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment-0003.obj>

From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:07:03 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:07:03 +0200
Subject: Source anaylsis
In-Reply-To: <20060725142236.lz6o47hwkkwo4sw4@webmail.unav.es>
Message-ID: <WED.26.JUL.2006.090703.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
> I have the same problem and I do not know how to fix it, even
> though I tried your solution with cfg.channelcmb.
>
> My case is : i have a set of 82 electrodes, I make freqanalysis
> with cfg.channelcmb = channelcombination({'all'
> 'all'},data.label);  thus I get
>
> freq =
>
>         label: {82x1 cell}
>        dimord: 'chan_freq'
>          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000
> 16.8000 17.2000 17.6000 18]
>     powspctrm: [82x11 double]
>      labelcmb: {3321x2 cell}
>     crsspctrm: [3321x11 double]
>           cfg: [1x1 struct]
>
> When I make sourceanalysis with a dipole as a reference and I have
> no problems but when I want to use a channel as a reference I
> always have the same error:
> The cross-spectral-density matrix is not complete.
>
> Can you help me?
>
> Best regards
> Juan

Dear Juan

Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
expect the latter, since computing coherence between one particular
EEG electrode and sources in the brain with DICS will not lead to
meaningfull results. The selected EEG electrode will pick up volume-
conducted activity from all sources in the brain (as opposed to an
EMG electrode) and would be referenced to the same reference
electrode as all other electrodes. So there will be a huge amount of
trivial coherence, whose spatial distribution in the brain will
probably more reflect the volume-conduction than something
physiologically interpretable.

Although I have not used DICS with a reference channel on EEG data, I
would expect the code to work if you follow a procedure like this:

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EEG electrodes
cfg.implicitref = name of implicit reference electrode used for the
EEG channels
cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
data_eeg = preprocessing(cfg)

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EMG electrodes, assuming that they are
bipolar channels
% cfg.reref may be used if you have two unipolar channels on the
muscle of interest
cfg.rectify = 'yes'   % this makes the spiky nature of the EMG more
visible, and improves the coherence estimate
cfg.baseline = 'yes'  % remove baseline prior to rectification, not
neccessary but still preferable
data_emg = preprocessing(cfg)

% combine EEG and EMG in one dataset, both are referenced differently
data = appenddata([], data_eeg, data_emg);

cfg = ... frequency analysis stuff
cfg.ouptu = 'powandcsd'
cfg.channelcmb = {'all', 'all'}
freq = freqanalysis(cfg, data);

cfg = ... source specific settings
cfg.refchan = one of your bipolar EMG channels
cfg.channel = all your EEG channels
source = sourceanalysis(cfg, freq)

best regards,
Robert


From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:46:59 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:46:59 +0200
Subject: import bva time-freq data
In-Reply-To: <LISTSERV%200607251739327460.3C55@NIC.SURFNET.NL>
Message-ID: <WED.26.JUL.2006.094659.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Stephan,

On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
> I try to import brainvision analyzer dat files of wavelet time-
> frequency
> decompositions into fieldtrip. I exported an average file, so it
> consists
> only of a single segment. However I have some problems with it I am
> not sure
...
> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
> \WT_all_ft_test.vmrk';
> cfg.trialdef.stim  = 'New Segment';
> %cfg.trialdef.segment  = 'yes';
> cfg.trialdef.timezero = 'yes';
> % cfg.trialdef.eventtype  = 'Time 0';

In general I would recommend to use the trialfun_general instead of
the trialfun_brainvision (but the latter should also work). The
trialfun_general was written later and will work for any dataformat
that is supported by read_fcdc_event, the trialfun_brainvision is
only there to support old scripts (i.e. scripts that predate the
read_fcdc_event function).

If not specified manually as trialfun, the trialfun_general or
trialfun_brainvision is selected automatically based on the cfg
settings. In your case you could do
    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
    cfg.trialdef.prestim    = number, latency in seconds (optional)
    cfg.trialdef.poststim   = number, latency in seconds (optional)
See teh help of DEFINETRIAL.

But in your case this is not the problem. The problem lies in
> cfg.trialdef.prestim    = 0.500;
> cfg.trialdef.poststim   = 1.600;

If I do
   event = read_fcdc_event('WT_all_ft_test.vmrk')
then I see that

 >> event(1)
ans =
         type: 'New Segment'
        value: '1'
       sample: 1
     duration: 0
       offset: []

 >> event(2)
ans =
         type: 'Time 0'
        value: '251'
       sample: 1
     duration: 0
       offset: []

The 'New Segment' event is at sample 1, hence you cannot select a
500ms pre-stimulus window before it (you cannot read before the
beginning of the data). The alternative 'Time 0' event is also at
sample 1. It specifies a value of 251, which surprises me. I would
expect that event to be present at sample 251.

Reading the marker file with a text editor, I see (abbreviated) the
explanation
    Mk<number>=<Type>,<Description>,<Position>,<Size>,
<Channelnumber>,<Date>
and the markers
    Mk1=New Segment,,1,1,0,00000000000000000000
    Mk2=Time 0,,251,1,0
So Mk2 should indeed be at sample 251.

With the matlab debugger in the private/read_event.m file (around
line 164) I notice that the problem lies in the missing value of the
second field, i.e.
    Mk2=Time 0,,251,1,0
is interpreted as
    Mk2=Time 0,251,1,0  (one comma less)
All fields are shifted by one, resulting in the sample number being
interpreted as the description (i.e. value).

I have fixed it in read_event, by replacing line 151 from
   tok = tokenize(line, '=');
into
   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
seperator
and by replacing line 157 from
   tok = tokenize(tok{2}, ',');
into
   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
the seperator
i.e. both calls to the tokenize function now use the third argument.
You can apply the same change to your copy, but please check that you
have an up to date version of tokenize that accepts 3 input arguments
(tokenize is in private as well). I will include the bug fix in the
upcoming nightly release of FieldTrip. Better upgrade tomorrow to the
latest FT version from the FTP server.

I think that sofar it was not noticed since people here at the
Donders typically use triggers as markers in their Brainvision data,
and in case of triggers the second field would not be empty.

If I now do
   cfg = [];
   cfg.dataset = 'WT_all_ft_test.vhdr'
   cfg.trialdef.eventtype = 'Time 0'
   cfg.trialdef.prestim  =  0.5000
   cfg.trialdef.poststim =  1
   cfg = definetrial(cfg)
I get
   evaluating trialfunction 'trialfun_general'
   found 2 events
   created 1 trials
   cfg =
        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: []
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
which is correct. The trial (cfg.trl) runs from sample 1 to sample
750 in the data file, and the first sample of that trial corresponds
with time cfg.trl(1,3)/fsample=-0.500 seconds.

Thanks for reporting the bug,
Robert


From stephan.bickel at ANATOM.UNIZH.CH  Wed Jul 26 17:28:54 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel, Anatomisches Inst.)
Date: Wed, 26 Jul 2006 11:28:54 -0400
Subject: import bva time-freq data
In-Reply-To: <30468A0B-0B5C-41EF-9FB3-84FD50036A22@fcdonders.ru.nl>
Message-ID: <WED.26.JUL.2006.112854.0400.FIELDTRIP@NIC.SURFNET.NL>

Hi Robert,

thank you very much for your response and for fixing the bug. Reading in the
trial information works fine now. Unfortunately I encounter another problem
when I want to read in the data with preprocessing. I wonder if it is yet
possible to read in time-frequency data from bva-dat files?

Sorry for bothering you again and thank you very much in advance.

Stephan


So when I do:

cfg = [];
cfg.dataset = 'WT_all_ft_test.vhdr'
cfg.trialdef.eventtype = 'Time 0'
cfg.trialdef.prestim = 0.5000
cfg.trialdef.poststim = 1
cfg = definetrial(cfg)
cfg.datafile   = 'WT_all_ft_test.dat';
cfg.channel = 'all';
[raw] = preprocessing(cfg)


I get:

cfg =

        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: 'WT_all_ft_test.dat'
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
        channel: 'all'

retaining exist trial definition
retaining exist event information
found 2 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
     dat(line,:) = str2num(str);

Error in ==> fieldtrip-20060725\private\read_data at 239
     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Error in ==> read_fcdc_data at 49
[dat] = read_data(varargin{:});

Error in ==> preprocessing at 366
   dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);



On Wed, 26 Jul 2006 09:46:59 +0200
  Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> wrote:
> Hi Stephan,
>
> On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
>> I try to import brainvision analyzer dat files of wavelet time-
>> frequency
>> decompositions into fieldtrip. I exported an average file, so it
>> consists
>> only of a single segment. However I have some problems with it I am
>> not sure
> ...
>> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
>> \WT_all_ft_test.vmrk';
>> cfg.trialdef.stim  = 'New Segment';
>> %cfg.trialdef.segment  = 'yes';
>> cfg.trialdef.timezero = 'yes';
>> % cfg.trialdef.eventtype  = 'Time 0';
>
> In general I would recommend to use the trialfun_general instead of  the
>trialfun_brainvision (but the latter should also work). The
> trialfun_general was written later and will work for any dataformat  that
>is supported by read_fcdc_event, the trialfun_brainvision is  only there to
>support old scripts (i.e. scripts that predate the  read_fcdc_event
>function).
>
> If not specified manually as trialfun, the trialfun_general or
> trialfun_brainvision is selected automatically based on the cfg  settings.
>In your case you could do
>    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
>    cfg.trialdef.prestim    = number, latency in seconds (optional)
>    cfg.trialdef.poststim   = number, latency in seconds (optional)
> See teh help of DEFINETRIAL.
>
> But in your case this is not the problem. The problem lies in
>> cfg.trialdef.prestim    = 0.500;
>> cfg.trialdef.poststim   = 1.600;
>
> If I do
>   event = read_fcdc_event('WT_all_ft_test.vmrk')
> then I see that
>
> >> event(1)
> ans =
>         type: 'New Segment'
>        value: '1'
>       sample: 1
>     duration: 0
>       offset: []
>
> >> event(2)
> ans =
>         type: 'Time 0'
>        value: '251'
>       sample: 1
>     duration: 0
>       offset: []
>
> The 'New Segment' event is at sample 1, hence you cannot select a  500ms
>pre-stimulus window before it (you cannot read before the  beginning of the
>data). The alternative 'Time 0' event is also at  sample 1. It specifies a
>value of 251, which surprises me. I would  expect that event to be present
>at sample 251.
>
> Reading the marker file with a text editor, I see (abbreviated) the
> explanation
>    Mk<number>=<Type>,<Description>,<Position>,<Size>,
> <Channelnumber>,<Date>
> and the markers
>    Mk1=New Segment,,1,1,0,00000000000000000000
>    Mk2=Time 0,,251,1,0
> So Mk2 should indeed be at sample 251.
>
> With the matlab debugger in the private/read_event.m file (around  line
>164) I notice that the problem lies in the missing value of the  second
>field, i.e.
>    Mk2=Time 0,,251,1,0
> is interpreted as
>    Mk2=Time 0,251,1,0  (one comma less)
> All fields are shifted by one, resulting in the sample number being
> interpreted as the description (i.e. value).
>
> I have fixed it in read_event, by replacing line 151 from
>   tok = tokenize(line, '=');
> into
>   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
> seperator
> and by replacing line 157 from
>   tok = tokenize(tok{2}, ',');
> into
>   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
> the seperator
> i.e. both calls to the tokenize function now use the third argument.  You
>can apply the same change to your copy, but please check that you  have an
>up to date version of tokenize that accepts 3 input arguments  (tokenize is
>in private as well). I will include the bug fix in the  upcoming nightly
>release of FieldTrip. Better upgrade tomorrow to the  latest FT version
>from the FTP server.
>
> I think that sofar it was not noticed since people here at the  Donders
>typically use triggers as markers in their Brainvision data,  and in case
>of triggers the second field would not be empty.
>
> If I now do
>   cfg = [];
>   cfg.dataset = 'WT_all_ft_test.vhdr'
>   cfg.trialdef.eventtype = 'Time 0'
>   cfg.trialdef.prestim  =  0.5000
>   cfg.trialdef.poststim =  1
>   cfg = definetrial(cfg)
> I get
>   evaluating trialfunction 'trialfun_general'
>   found 2 events
>   created 1 trials
>   cfg =
>        dataset: 'WT_all_ft_test.vhdr'
>       trialdef: [1x1 struct]
>       datafile: []
>     headerfile: 'WT_all_ft_test.vhdr'
>       trialfun: 'trialfun_general'
>          event: [1x2 struct]
>            trl: [1 750 -250]
>        version: [1x1 struct]
> which is correct. The trial (cfg.trl) runs from sample 1 to sample  750 in
>the data file, and the first sample of that trial corresponds  with time
>cfg.trl(1,3)/fsample=-0.500 seconds.
>
> Thanks for reporting the bug,
> Robert


From r.oostenveld at FCDONDERS.RU.NL  Thu Jul 27 09:26:26 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Thu, 27 Jul 2006 09:26:26 +0200
Subject: import bva time-freq data
In-Reply-To: <web-4062129@idmailbe2.unizh.ch>
Message-ID: <THU.27.JUL.2006.092626.0200.FIELDTRIP@NIC.SURFNET.NL>

On 26 Jul 2006, at 17:28, Stephan Bickel, Anatomisches Inst. wrote:
> thank you very much for your response and for fixing the bug.
> Reading in the trial information works fine now. Unfortunately I
> encounter another problem when I want to read in the data with
> preprocessing. I wonder if it is yet possible to read in time-
> frequency data from bva-dat files?

...

> Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
>     dat(line,:) = str2num(str);
>
> Error in ==> fieldtrip-20060725\private\read_data at 239
>     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Hi Stephan,

I have no idea how time-frequency data is stored in the BVA *.dat
files. Even if the low-level reading function were to support it, the
PREPROCESSING function at least would not be able to deal with it,
since the data structure returned by preprocessing does not allow for
3D data (channelXtimeXfrequency). That type of data would have to be
stored in a structure that is compatible with the output of
FREQANALYSIS. The preprocessing function is made for raw continuous
(or trial-based) data, and it happens to work as well for averaged
ERP data, although it is not specifically made for that.

For other external software (BESA, EEGLAB, LORETA) I have created
xxx2fieldtrip functions, i.e. functions that import already processed
data and format it into the fieldtrip structure that is most
compatible with it. In this case I can imagine a
BRAINVISION2FIELDTRIP function that would read your dat file and
return a "freq" structure. Besides making a brainvision2fieldtrip
function, it requires making either the read_brainvision_eeg smarter
(or probably making a seperate read_brainvision_tfr function) and
improving the detection of teh content of the BVA *.dat file. Please
have a look in besa2fieldtrip to see how I approach it there.

best regards,
Robert


From jciveira at UNAV.ES  Thu Jul 27 18:54:27 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Thu, 27 Jul 2006 18:54:27 +0200
Subject: Source anaylsis
In-Reply-To: <9A0AA0F8-A8AD-4958-B224-4E3547238053@FCDONDERS.RU.NL>
Message-ID: <THU.27.JUL.2006.185427.0200.FIELDTRIP@NIC.SURFNET.NL>

   Dear Robert:

   Thanks for your answer but my problems are still there. The main
problem is that i can not use DICS with a reference channel on EEG
data, no matter if  the reference is an EEG channel or EMG or some
other channel.

   I have taken my measures with 85 electrodes, some of them are
EEG1010, some EOG and some other that I have placed myself.

   I do the freq analysis and then the source analysis. Then I can
make another source analysis with refdip but there is always errors
when I use refchan. I do not think is a problem of the reference ,
cos always the program says that the  cross-spectral-density with
the reference channel is not complete .

   I have done the freq analysis making cfg.channelcmb combinations of
all channels with only one, so i make sure that the
cross-spectral-density is complete,but when I make source analysis
with that channel as a reference is the same.

   Searching in the code I have come up with the fact that the error
is always the condition in line 195 of the function
prepare_freq_matrices (version 2006/03/08). My idea is that there is
a confussion because if we have a set of N channels and we make  
combinations with channelcombination, the result is N-1 combinations  
because we do not include the combination of the reference with itself.

I hope i explain myself clear enough, anyway I am trying to fix the  
problem by myself but any help or suggestion will be great

Juan


Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
>> I have the same problem and I do not know how to fix it, even   
>> though I tried your solution with cfg.channelcmb.
>>
>> My case is : i have a set of 82 electrodes, I make freqanalysis   
>> with cfg.channelcmb = channelcombination({'all'
'all'},data.label);
>>  thus I get
>>
>> freq =
>>
>>        label: {82x1 cell}
>>       dimord: 'chan_freq'
>>         freq: [14 14.4000 14.8000 15.2000 15.6000 16
16.4000
>> 16.8000 17.2000 17.6000 18]
>>    powspctrm: [82x11 double]
>>     labelcmb: {3321x2 cell}
>>    crsspctrm: [3321x11 double]
>>          cfg: [1x1 struct]
>>
>> When I make sourceanalysis with a dipole as a reference and I
have
>> no problems but when I want to use a channel as a reference I   
>> always have the same error:
>> The cross-spectral-density matrix is not complete.
>>
>> Can you help me?
>>
>> Best regards
>> Juan
>
> Dear Juan
>
> Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
> expect the latter, since computing coherence between one particular
EEG
> electrode and sources in the brain with DICS will not lead to
> meaningfull results. The selected EEG electrode will pick up
> volume-conducted activity from all sources in the brain (as opposed
to
> an EMG electrode) and would be referenced to the same reference
> electrode as all other electrodes. So there will be a huge amount
of
> trivial coherence, whose spatial distribution in the brain will
> probably more reflect the volume-conduction than something
> physiologically interpretable.
>
> Although I have not used DICS with a reference channel on EEG data,
I
> would expect the code to work if you follow a procedure like this:
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EEG electrodes
> cfg.implicitref = name of implicit reference electrode used for the
EEG
> channels
> cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
> data_eeg = preprocessing(cfg)
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EMG electrodes, assuming that they are
> bipolar channels
> % cfg.reref may be used if you have two unipolar channels on the
muscle
> of interest
> cfg.rectify = 'yes'   % this makes the spiky nature of the EMG
more
> visible, and improves the coherence estimate
> cfg.baseline = 'yes'  % remove baseline prior to rectification,
not
> neccessary but still preferable
> data_emg = preprocessing(cfg)
>
> % combine EEG and EMG in one dataset, both are referenced
differently
> data = appenddata([], data_eeg, data_emg);
>
> cfg = ... frequency analysis stuff
> cfg.ouptu = 'powandcsd'
> cfg.channelcmb = {'all', 'all'}
> freq = freqanalysis(cfg, data);
>
> cfg = ... source specific settings
> cfg.refchan = one of your bipolar EMG channels
> cfg.channel = all your EEG channels
> source = sourceanalysis(cfg, freq)
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 09:49:55 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 09:49:55 +0200
Subject: Source anaylsis
In-Reply-To: <20060727185427.3r2gdy6zkk0o4og8@webmail.unav.es>
Message-ID: <FRI.28.JUL.2006.094955.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Juan

>   I do the freq analysis and then the source analysis. Then I can
> make another source analysis with refdip but there is always errors
> when I use refchan. I do not think is a problem of the reference ,
> cos always the program says that the  cross-spectral-density with
> the reference channel is not complete .
>
>   I have done the freq analysis making cfg.channelcmb combinations of
> all channels with only one, so i make sure that the
> cross-spectral-density is complete,but when I make source analysis
> with that channel as a reference is the same.
>
>   Searching in the code I have come up with the fact that the error
> is always the condition in line 195 of the function
> prepare_freq_matrices (version 2006/03/08). My idea is that there is
> a confussion because if we have a set of N channels and we make
> combinations with channelcombination, the result is N-1
> combinations because we do not include the combination of the
> reference with itself.

Say for simplicity that you want EMG as refchan. In freqanalysis,
cfg.channelcmb should be all with all. You need the CSD between all
EEG channels (all EEG with all EEG), and of all EEG channels with the
EMG. In prepare_freq_matrices the CSD matrix (Cf) of EEG with EEG is
made (Neeg X Neeg), the CSD vector (Cr) of all EEG channels with the
reference EMG channel (Neeg X 1), and the power of the reference
channel (Pr). Together they could also be shaped in a (Neeg+1) X (Neeg
+1) matrix in principle, but that is not done for DICS.

I have tried it out myself, and cannot reproduce your problem. Please
find my test script below. It uses some simulated EEG data. The
beginning of the script is copied from http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:compute_forward_simulated_data_and_apply_a_di
pole_fit

The initial part generates raw EEG data, then I add a EMG channel
(with noise, but it is only to demonstrate that fieldtrip does not
give an error), then I do freqanalysis and sourceanalysis. Please try
out and compare my cfgs with yours, since I suspect your
configuration to be the problem.

best regards,
Robert


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% create a set of electrodes, randomly placed on the sphere
elec = [];
elec.pnt = randn(128,3);
dum = sqrt(sum(elec.pnt.^2,2));
elec.pnt = elec.pnt ./ [dum dum dum];  % scale them to a unit sphere
for i=1:128
    elec.label{i} = sprintf('%03d', i);
end

% create a concentric 3-sphere volume conductor, the radius is the
same as for the electrodes
vol = [];
vol.r = [0.88 0.92 1.00]; % radii of spheres
vol.c = [1 1/80 1];       % conductivity
vol.o = [0 0 0];          % center of sphere

% create a dipole simulation with one dipole and a 10Hz sine wave
cfg      = [];
cfg.vol  = vol;             % see above
cfg.elec = elec;            % see above
cfg.dip.pos = [0 0.5 0.3];
cfg.dip.mom = [1 0 0]';     % note, it should be transposed
cfg.dip.frequency = 10;
cfg.ntrials = 10;
cfg.triallength = 1;        % seconds
cfg.fsample = 250;          % Hz
raw1 = dipolesimulation(cfg);

using headmodel specified in the configuration
using electrodes specified in the configuration
selected 128 electrodes
computing simulated data
RMS value of simulated data is 0.0746553

%%%%%%%%%%%%%%%%%%%%%%%%
%%%%% Here it starts deviating from the demo script on the wiki %%%%
%%%%%%%%%%%%%%%%%%%%%%%%

% add a fake EMG channel, containing noise
for i=1:10
   raw1.trial{i}(end+1,:) = randn(1,250);
end
raw1.label{end+1} = 'EMG'

raw1 =

       trial: {1x10 cell}
        time: {1x10 cell}
        elec: [1x1 struct]
     fsample: 250
       label: {1x129 cell}
         cfg: [1x1 struct]

cfg = []
cfg.method = 'mtmfft';
cfg.taper = 'hanning';
cfg.foilim = [2 20];
cfg.output = 'powandcsd';
cfg.channelcmb = {'all', 'all'}

cfg =

         method: 'mtmfft'
          taper: 'hanning'
         foilim: [2 20]
         output: 'powandcsd'
     channelcmb: {'all'  'all'}

freq1 = freqanalysis(cfg, raw1)
WARNING: using only one taper for specified smoothing
processing trial 1, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 2, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 3, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 4, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 5, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 6, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 7, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 8, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 9, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 10, nfft: 250 samples, taper length: 250 samples, 1
tapers

freq1 =
          label: {1x129 cell}
        dimord: 'chan_freq'
          freq: [2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20]
     powspctrm: [129x19 double]
      labelcmb: {8256x2 cell}
     crsspctrm: [8256x19 double]
          elec: [1x1 struct]
           cfg: [1x1 struct]

cfg = [];
cfg.vol = vol;
cfg.freq = 10;
cfg.frequency = 10;
cfg.method = 'dics';
cfg.refchan = 'EMG';
cfg.grid.resolution = 0.3

cfg =
           vol: [1x1 struct]
          freq: 10
     frequency: 10
        method: 'dics'
       refchan: 'EMG'
          grid: [1x1 struct]

source1 = sourceanalysis(cfg, freq1)
using headmodel specified in the configuration
using electrodes specified in the data
selected 128 electrodes
106 dipoles inside, 237 dipoles outside brain
making tight grid
106 dipoles inside, 110 dipoles outside brain
scanning repetition 1
scanning grid
Warning: cross-spectral density matrix is rank deficient
 > In <a href="error:/Users/roberto/matlab/forwinv/beamformer.m,
270,1">beamformer at 270</a>
   In <a href="error:/Users/roberto/matlab/fieldtrip/sourceanalysis.m,
806,1">sourceanalysis at 806</a>
total time in sourceanalysis 75.4 seconds

source1 =
         xgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         ygrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         zgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
           dim: [6 6 6]
           vol: [1x1 struct]
     frequency: 10
           pos: [216x3 double]
        inside: [1x106 double]
       outside: [1x110 double]
        method: 'average'
           avg: [1x1 struct]
           cfg: [1x1 struct]

diary off


From floris.delange at FCDONDERS.RU.NL  Fri Jul 28 13:01:00 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Fri, 28 Jul 2006 13:01:00 +0200
Subject: sourcestatistics: what to compare?
Message-ID: <FRI.28.JUL.2006.130100.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Fieldtrippers,

I have a question about what conditions/comparisons make most sense to
statistically compare with sourcestatistics.
I have 2 conditions, A and B, and each condtion has its own pre-stimulus
baseline: basA and basB.

Now, first I'd like to see which regions are showing stronger beta-band
power during A than the baseline. So I want to compare A to basA.
There's 2 ways of going about it:
1) compare (the grand average of) A.avg.pow with basA.avg.pow
2) compare (the grand average of) the neural activity index (NAI) of A
with the baseline

Any ideas which would be best? Or is this just an empirical issue?
(i.e., take the best)

To compare A and B, one has even more options:
1) compare A.avg.pow with B.avg.pow
2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
3) compare NAI(A) with NAI(B)
4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)

Is there's any principled reason for a priori picking any of the above
choices?

Your input is greatly appreciated,
Kind regards,

Floris


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 16:17:24 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 16:17:24 +0200
Subject: sourcestatistics: what to compare?
In-Reply-To: <44C9EE6C.8030003@fcdonders.ru.nl>
Message-ID: <FRI.28.JUL.2006.161724.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris

On 28 Jul 2006, at 13:01, Floris de Lange wrote:
> I have a question about what conditions/comparisons make most sense
> to statistically compare with sourcestatistics.
> I have 2 conditions, A and B, and each condtion has its own pre-
> stimulus baseline: basA and basB.
>
> Now, first I'd like to see which regions are showing stronger beta-
> band power during A than the baseline. So I want to compare A to
> basA. There's 2 ways of going about it:
> 1) compare (the grand average of) A.avg.pow with basA.avg.pow
> 2) compare (the grand average of) the neural activity index (NAI)
> of A with the baseline
>
> Any ideas which would be best? Or is this just an empirical issue?
> (i.e., take the best)

Both are valid, you could take the best. There are some exceptions/
predictions though that I can share with you: the NAI is power
divided by estimated (projected) noise. It is inherently difficult to
estimate the noise level (the default is the smallest singular value
of the CSD matrix, but you can manipulate it with cfg.lambda in
sourceanalysis). If you have different numbers of trials in active
v.s. baseline, the noise estimate probably will be different. So then
you would have NAIa = Pa/Na and NAIb = Pb/Nb, if Pa==Pb and Na~=Nb,
these will be trivially different, which is not what you want
(remember Pa==Pb was assumed).

You can compare the projected noise in both conditions, the source-
noise-estimate is uniformely scaled (i.e. on all voxels the same)
with the estimated noise from the CSD. There is no reason to assume
that the noise is different, so if the noise is different, you would
have to explain that.

> To compare A and B, one has even more options:
> 1) compare A.avg.pow with B.avg.pow
> 2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
> 3) compare NAI(A) with NAI(B)
> 4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)
>
> Is there's any principled reason for a priori picking any of the
> above choices?

I presume that by "compare" you mean "look at the the difference".

In general you should only use the NAI for plotting data if you only
have one condition (and no baseline). The noise estimate that goes
into the NAI is too poor to be of real value. Although sometines it
looks nice in figures, the NAI tends not to be very robust (it
contains global flucuations that mess up the statistic, although the
figure still looks nice).

The baseline is only interesting if you expect something differently
to happen in the two baselines. If that is the case, you better try
to get another baseline in your experimental design. If you would
combine actA, basA, actB, basB into a single number, and that number
would be significant, you still would not be able to tell whether the
difference is betwen the active conditions, theis baselines, or some
weird combination of the 4 "conditions" that you combine.

Typically there is no reason to assume taht the baselines are
different. So
   (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
= (A.avg.pow/basC.avg.pow) with (B.avg.pow/basC.avg.pow), where C
means common
= (A.avg.pow-B.avg.pow)/basC.avg.pow.
I.e. you are comparing A and B directly, with a spatially normalizing
for the different baseline-estimated noise at different source
locations.

Whether you want to spatially normalize (divide by some baseline
estimated-source-power) depends on the leadfields (it is neccessary
if cfg.normalize=no, if you want to do multiple comparison
correction, and if you want to look at an additive effect). If you
want to look at a multiplicative effect, you would look at the ratio
between two conditions instead of the difference. The ratio of the
conditions is best implemented with a log transform, after which you
again can look at the difference between the log-transformed power
estimates in both conditions (log(A/B) = log(A) - log(B)). In the
ratio or log-ratio between the powers in both conditions you have
also corrected for the spatial (depth) noise-bias.

good luck with the stats,
Robert

PS as I am on holiday the next 2 weeks, you could ask Ingrid about
more details on this. She is more or less at the same stage in her
analysis (about to do group sourcestatistics), but I suspect that she
already has more "feel" for these issues. Or ask JM.




From d.vanbarneveld at STUDENT.SCIENCE.RU.NL  Thu Jul  6 16:58:40 2006
From: d.vanbarneveld at STUDENT.SCIENCE.RU.NL (Denise van Barneveld)
Date: Thu, 6 Jul 2006 16:58:40 +0200
Subject: problem with text in topoplot and triplot
Message-ID: <THU.6.JUL.2006.165840.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear all,

since I started using the new version of Matlab (R2006a) on a mac, I
have this problem with text in the figures which I generate with
topoplot.m and triplot.m. The upper half of the text lines (title,
colorbar, electrodes, etc) is not visible (see attached figure).
Does anyone have this problem too? Especially people who are using
this new version on a pc. And does anyone perhaps has a solution for
this problem? I am currently looking in the code, if I can find where
it goes wrong, but I wonder whether it is a problem only with the new
version, or only at a mac, or only at my computer.

Best regards,
Denise van Barneveld

-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: application/applefile
Size: 48798 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment-0002.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: topo.jpg
Type: image/jpeg
Size: 85086 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060706/ed07759c/attachment-0002.jpg>

From muthuraman10 at HOTMAIL.COM  Fri Jul  7 12:04:26 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Fri, 7 Jul 2006 10:04:26 +0000
Subject: query on confidence limit, DICS
In-Reply-To: <37FAA4DD-5DD8-4F2F-ABA6-D9F47F6AB95E@fcdonders.ru.nl>
Message-ID: <FRI.7.JUL.2006.100426.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

As i am able to sucessfully use the coherence analysis using freq_mtmfft and
frq_mtmwelch to calculate the coherence
i would like to know is there a way to calculate the confidence limit for
the coherence calculated from the freqanalysis, is there a standard code

Next, i would like to follow with the source analysis 'DICS' is there a
tutorial available

thanking you

with regards
Muthuraman


>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on CTF software!!!
>Date: Fri, 23 Jun 2006 15:23:22 +0200
>
>dear Muthuraman
>
>On 23 Jun 2006, at 12:54, Muthuraman Muthuraman wrote:
>>For the Beamformer source analysis, do we need the CTF software
>>because when i am trying out the tutorial in beamforming
>
>You do not need the ctf software to follow the tutorial. The tutorial  is
>based however on a ctf MEG dataset.
>
>>CTF_READ_RES4 [v  1.12]
>>....done (  0.67 sec)
>>
>>i get the error
>>
>>??? Error using ==> read_fcdc_header
>>could not read CTF res4 header file
>>
>>Error in ==> preprocessing at 263
>>hdr = read_fcdc_header(cfg.headerfile);
>
>Based on your error I suspect that matlab/fieldtrip cannot find the
>dataset that you are trying to analyze, or that you matlab path is
>incorrect. The function "CTF_READ_RES4" is not supposed to be called  (it
>is part of the NIH ctf import routines, e.g. used in EEGLAB),  since the
>default private/read_ctf_res4.m function can deal with the  tutorial
>dataset just fine. Please look into the code of  read_fcdc_header (using
>the matlab debugger) to figure out what your  exact problem is.
>
>best
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul  7 12:36:02 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 7 Jul 2006 12:36:02 +0200
Subject: query on confidence limit, DICS
In-Reply-To: <BAY121-F19F77F86077022EC6849B3CE740@phx.gbl>
Message-ID: <FRI.7.JUL.2006.123602.0200.FIELDTRIP@NIC.SURFNET.NL>

On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
> As i am able to sucessfully use the coherence analysis using
> freq_mtmfft and frq_mtmwelch to calculate the coherence
> i would like to know is there a way to calculate the confidence
> limit for the coherence calculated from the freqanalysis, is there
> a standard code

You can compute a jackknife estimate of the variance in the coherence
using the freqdescriptives function. That requires keeptrials=yes in
freqanalysis.

> Next, i would like to follow with the source analysis 'DICS' is
> there a tutorial available

No there is no tutorial available specifically for that. You can look
at the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
options are cfg.method=dics, cfg.refchan and/or cfg.refdip.

Robert


From j.poort at NIN.KNAW.NL  Tue Jul 11 11:43:49 2006
From: j.poort at NIN.KNAW.NL (Jasper Poort)
Date: Tue, 11 Jul 2006 11:43:49 +0200
Subject: Confidence intervals for evoked potentials
Message-ID: <TUE.11.JUL.2006.114349.0200.FIELDTRIP@NIC.SURFNET.NL>

 
Hi all, I have some questions regarding confidence intervals for timelocked data in FieldTrip. I noted that a field .var is returned by timelockanalysis

line 510 timelockanalysis
% compute the variance
tmp1 = repmat(dofvec(:)', [nchan 1]);
tmp2 = repmat(dofvec(:)', [nchan 1])-1;
var = (ss - (s.^2)./tmp1) ./ tmp2;

Assuming that the sampling distribution is normal I can presumably use this field for computation of confidence intervals of the .avg field? 
Is nonparametric estimation of variance using a bootstrap or jackknife also implemented somewhere? In general I'm just curious to know how others compute confidence intervals for evoked potentials in FieldTrip, for example to get an idea at what point in time two conditions start to differ, 

best regards, Jasper
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060711/806a2020/attachment-0002.html>

From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 15:50:09 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 15:50:09 +0200
Subject: Confidence intervals for evoked potentials
In-Reply-To: <017d01c6a4ce$835853e0$910b57c0@ABSYS>
Message-ID: <TUE.11.JUL.2006.155009.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Jasper,

On 11 Jul 2006, at 11:43, Jasper Poort wrote:
> Hi all, I have some questions regarding confidence intervals for
> timelocked data in FieldTrip. I noted that a field .var is returned
> by timelockanalysis
>
> line 510 timelockanalysis
> % compute the variance
> tmp1 = repmat(dofvec(:)', [nchan 1]);
> tmp2 = repmat(dofvec(:)', [nchan 1])-1;
> var = (ss - (s.^2)./tmp1) ./ tmp2;
>
> Assuming that the sampling distribution is normal I can presumably
> use this field for computation of confidence intervals of the .avg
> field?

Yes, you can use it to compute the channel-timepoint specific
standard-error of mean (SEM).

> Is nonparametric estimation of variance using a bootstrap or
> jackknife also implemented somewhere? In general I'm just curious
> to know how others compute confidence intervals for evoked
> potentials in FieldTrip, for example to get an idea at what point
> in time two conditions start to differ,

Bootstrap and jackknife estimates of variance are only different from
teh normal variance estimate if the measure involves a nonlinear or
biased estimate (i.e. coherence). The jackknife estimate of variance
of the ERP is exactly the same as the normal estimate of variance,
since the ERP is a plain (linear) average. The same holds for
bootstrap estimates.

So that means that for ERPs the jackknife and bootstrap only would be
a computationally expensive methods, resulting in the same variance
as usual. Therefore they are not implemented.

best regards,
Robert


From muthuraman10 at HOTMAIL.COM  Tue Jul 11 16:59:58 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 11 Jul 2006 14:59:58 +0000
Subject: DICS
In-Reply-To: <E2FC4366-C6DD-476C-B902-6EBF864D6B48@fcdonders.ru.nl>
Message-ID: <TUE.11.JUL.2006.145958.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thanks for the suggestion to use the jackknife variance estimate it works
thanks a lot

To use the DICS method

We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG channels
recording
and also we have the 64 channel measurement in FRMI

the electrode locations we have it in a file with 3 columns x,y,z for each
electrode
using the polhemus sensors but our own software

In the Source analysis
Cfg.method='DICS';
but Cfg.xgrid, ygrid and zgrid
the grid values are the sensor location values

or it should be the
cfg.grid.pos-where i can give in the 3 values [x y z] for each electrode

I am a starter in the source analysis in combining EEG-FMRI
please let me know

with regards
muthuraman.






>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] query on confidence limit, DICS
>Date: Fri, 7 Jul 2006 12:36:02 +0200
>
>On 7 Jul 2006, at 12:04, Muthuraman Muthuraman wrote:
>>As i am able to sucessfully use the coherence analysis using  freq_mtmfft
>>and frq_mtmwelch to calculate the coherence
>>i would like to know is there a way to calculate the confidence  limit for
>>the coherence calculated from the freqanalysis, is there  a standard code
>
>You can compute a jackknife estimate of the variance in the coherence
>using the freqdescriptives function. That requires keeptrials=yes in
>freqanalysis.
>
>>Next, i would like to follow with the source analysis 'DICS' is  there a
>>tutorial available
>
>No there is no tutorial available specifically for that. You can look  at
>the beamformer tutorial at http://www2.ru.nl/fcdonders/fieldtrip/
>doku.php?id=fieldtrip:documentation:tutorial:beamformer, the relevant
>options are cfg.method=dics, cfg.refchan and/or cfg.refdip.
>
>Robert


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 11 17:51:28 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 11 Jul 2006 17:51:28 +0200
Subject: DICS
In-Reply-To: <BAY121-F26D094D1C667F7EAEC07B0CE680@phx.gbl>
Message-ID: <TUE.11.JUL.2006.175128.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman

On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
> To use the DICS method
>
> We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG
> channels recording
> and also we have the 64 channel measurement in FRMI
>
> the electrode locations we have it in a file with 3 columns x,y,z
> for each electrode
> using the polhemus sensors but our own software
>
> In the Source analysis
> Cfg.method='DICS';
> but Cfg.xgrid, ygrid and zgrid
> the grid values are the sensor location values
>
> or it should be the
> cfg.grid.pos-where i can give in the 3 values [x y z] for each
> electrode

in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the
locations at which the _dipoles_ will be placed, it is not the
position of the electrodes. At each grid location a spatial filter is
constructed and the power (and coherence) is computed. If you do not
want to scan on a full 3-D grid, but on only a few dipole locations
of interest, you can use cfg.grid.pos to specify those dipole locations.

The electrode positions are specified in the cfg.elec structure, or
in the cfg.elecfile filename. You can find more details on the
definition of electrode positions at http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
_magnetometers_or_gradiometers_described

Furthermore, you can try using the READ_FCDC_ELEC function for
reading the electrode locations, but probably taht will not work
since youy are using your own software for digitization.
Alternatively, you can manually construct an "elec" electrode
definition.

best regards
Robert

PS please upgrade to a recent fieldtrip version, the sourceanalysis
options cfg.xgrid/ygrid/zgrid have been renamed. The old options
still will work, but in general it is preferable to work with a
recent copy of the toolbox.


From muthuraman10 at HOTMAIL.COM  Thu Jul 13 13:11:35 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Thu, 13 Jul 2006 11:11:35 +0000
Subject: DICS
In-Reply-To: <43B3E5A4-716B-48A0-8CE0-5F8810DC4A65@fcdonders.ru.nl>
Message-ID: <THU.13.JUL.2006.111135.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for the suggestions
I have dowloaded the latest version now and i am working with it
i have created the own elec file which is "elec1" and these are the steps i
follow
please go through it and reply me, why is the error

-First i calculate the coherence using the freqanalysis and freqdescriptives
which works perfectly
-Second i calculate the Grid using Prepare_leadfield

vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1);

which also works i get the output for the Grid

-Third I want use the source analysis for the DICS

cfg.method='dics';
cfg.grid=grid;
cfg.elecfile=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
??? Error using ==> fprintf
Function is not defined for 'struct' inputs.

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
  fprintf('reading electrodes from file %s\n', cfg.elecfile);

Error in ==> sourceanalysis at 574
[vol, sens, cfg] = prepare_vol_sens(cfg, data);




-I also tried grid using the freqmtmfft-which as the results for the
freqanalysis
vol   = [];
vol.r = 12 * [0.88 0.92 1.00];
vol.c = [1 1/80 1];
vol.o = [0 0 0];
>>cfg1            = [];
cfg1.elec       = elec1;
cfg1.vol        = vol;
cfg1.resolution = 2;
grid = prepare_leadfield(cfg1,freqmtmfft);

I get error

using headmodel specified in the configuration
using electrodes specified in the configuration
??? Error using ==> delaunayn
Not enough unique points to do tessellation.

Error in ==> delaunay at 49
    tri = delaunayn([x(:) y(:)]);

Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));

Error in ==> prepare_leadfield at 179
[vol, sens, cfg] = prepare_vol_sens(cfg, data);


Thanking you

with regards
Muthuraman.




>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Tue, 11 Jul 2006 17:51:28 +0200
>
>Dear Muthuraman
>
>On 11 Jul 2006, at 16:59, Muthuraman Muthuraman wrote:
>>To use the DICS method
>>
>>We have a 64 channel EEG measurement from (Neuroscan) and 4 EMG  channels
>>recording
>>and also we have the 64 channel measurement in FRMI
>>
>>the electrode locations we have it in a file with 3 columns x,y,z  for
>>each electrode
>>using the polhemus sensors but our own software
>>
>>In the Source analysis
>>Cfg.method='DICS';
>>but Cfg.xgrid, ygrid and zgrid
>>the grid values are the sensor location values
>>
>>or it should be the
>>cfg.grid.pos-where i can give in the 3 values [x y z] for each  electrode
>
>in SOURCEANALYSIS you specify using cfg.xgrid/ygrid/zgrid the  locations at
>which the _dipoles_ will be placed, it is not the  position of the
>electrodes. At each grid location a spatial filter is  constructed and the
>power (and coherence) is computed. If you do not  want to scan on a full
>3-D grid, but on only a few dipole locations  of interest, you can use
>cfg.grid.pos to specify those dipole locations.
>
>The electrode positions are specified in the cfg.elec structure, or  in the
>cfg.elecfile filename. You can find more details on the  definition of
>electrode positions at http://www2.ru.nl/fcdonders/ fieldtrip/doku.php?
>id=fieldtrip:documentation:frequently_asked_questions#how_are_electrodes
>_magnetometers_or_gradiometers_described
>
>Furthermore, you can try using the READ_FCDC_ELEC function for  reading the
>electrode locations, but probably taht will not work  since youy are using
>your own software for digitization.  Alternatively, you can manually
>construct an "elec" electrode  definition.
>
>best regards
>Robert
>
>PS please upgrade to a recent fieldtrip version, the sourceanalysis
>options cfg.xgrid/ygrid/zgrid have been renamed. The old options  still
>will work, but in general it is preferable to work with a  recent copy of
>the toolbox.


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 14 11:37:47 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 14 Jul 2006 11:37:47 +0200
Subject: DICS
In-Reply-To: <BAY121-F32B67944419449CAB9EEF6CE6E0@phx.gbl>
Message-ID: <FRI.14.JUL.2006.113747.0200.FIELDTRIP@NIC.SURFNET.NL>

On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:

> cfg.method='dics';
> cfg.grid=grid;
> cfg.elecfile=elec1;
> cfg.vol=vol;
> [source]=sourceanalysis(cfg,freqmtmfft);
> using headmodel specified in the configuration
> ??? Error using ==> fprintf
> Function is not defined for 'struct' inputs.
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>  fprintf('reading electrodes from file %s\n', cfg.elecfile);

You should specify either cfg.elecfile with a string (filename) or
cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)

> I get error
>
> using headmodel specified in the configuration
> using electrodes specified in the configuration
> ??? Error using ==> delaunayn
> Not enough unique points to do tessellation.
>
> Error in ==> delaunay at 49
>    tri = delaunayn([x(:) y(:)]);
>
> Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>
> Error in ==> prepare_leadfield at 179
> [vol, sens, cfg] = prepare_vol_sens(cfg, data);

That indicates that less than 3 electrodes were selected. Probably
there is a mismatch between the channel names in the data and the
names of the electrodes. FT selects the channels and electrodes based
on the corresponding labels, if they do not correspond then no
channels+electrodes are selected.

Robert


From muthuraman10 at HOTMAIL.COM  Mon Jul 17 16:12:14 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Mon, 17 Jul 2006 14:12:14 +0000
Subject: DICS
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <MON.17.JUL.2006.141214.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
but not when using the mtmwelch is this a problem in the mtmwelch
or i am doing something wrong

Please go through it

thanking you

with regards
muthuraman



>From: Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL>
>Reply-To: FieldTrip discussion list <FIELDTRIP at NIC.SURFNET.NL>
>To: FIELDTRIP at NIC.SURFNET.NL
>Subject: Re: [FIELDTRIP] DICS
>Date: Fri, 14 Jul 2006 11:37:47 +0200
>
>On 13 Jul 2006, at 13:11, Muthuraman Muthuraman wrote:
>
>>cfg.method='dics';
>>cfg.grid=grid;
>>cfg.elecfile=elec1;
>>cfg.vol=vol;
>>[source]=sourceanalysis(cfg,freqmtmfft);
>>using headmodel specified in the configuration
>>??? Error using ==> fprintf
>>Function is not defined for 'struct' inputs.
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 310
>>  fprintf('reading electrodes from file %s\n', cfg.elecfile);
>
>You should specify either cfg.elecfile with a string (filename) or
>cfg.elec with a structure (i.e. cfg.elec.pnt and cfg.elec.label)
>
>>I get error
>>
>>using headmodel specified in the configuration
>>using electrodes specified in the configuration
>>??? Error using ==> delaunayn
>>Not enough unique points to do tessellation.
>>
>>Error in ==> delaunay at 49
>>    tri = delaunayn([x(:) y(:)]);
>>
>>Error in ==> fieldtrip-20060712\private\prepare_vol_sens at 330
>>  elec.tri   = delaunay(elec.prj(:,1), elec.prj(:,2));
>>
>>Error in ==> prepare_leadfield at 179
>>[vol, sens, cfg] = prepare_vol_sens(cfg, data);
>
>That indicates that less than 3 electrodes were selected. Probably  there
>is a mismatch between the channel names in the data and the  names of the
>electrodes. FT selects the channels and electrodes based  on the
>corresponding labels, if they do not correspond then no
>channels+electrodes are selected.
>
>Robert


From floris.delange at FCDONDERS.RU.NL  Tue Jul 25 10:24:18 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Tue, 25 Jul 2006 10:24:18 +0200
Subject: sourcestatistics question
Message-ID: <TUE.25.JUL.2006.102418.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear FieldTrippers,

I have a question concerning sourcestatistics.
I have calculated the source localization using DICS for 11 individual
subjects. Now I would like to test for the significance of the sources
using sourcestatistics. The examples I find in the tutorial are about
checking for significance within one subject by shuffling the individual
trials. However, what I would like to do is to do the statistics on the
2nd level (subject), and to test for the consistency of the source
reconstructions across subjects.
Is this possible with sourcestatistics? If so, how should I feed the
data to sourcestatistics? If I say something like stats =
sourcestatistics(cfg,sources{1:11}) where each source{} is a normalized
source (calculated by sourceanalysis, normalized with
sourceinterpolate), it complains (rightfully) that there's a reference
to the non-existing field 'statistic'.

Any ideas are greatly appreciated!

Kind regards,
Floris


From muthuraman10 at HOTMAIL.COM  Tue Jul 25 10:34:30 2006
From: muthuraman10 at HOTMAIL.COM (Muthuraman Muthuraman)
Date: Tue, 25 Jul 2006 08:34:30 +0000
Subject: Source anaylsis
In-Reply-To: <242E1EA4-D9FE-42D8-B23A-B689CBBF7FA8@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.083430.0000.FIELDTRIP@NIC.SURFNET.NL>

Hello

Thankyou for pointing out errors
now i am able to calculate the Grid values
but when i wanted to use the source analysis for the freqmtmwelch

cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmwelch);
using headmodel specified in the configuration
using electrodes specified in the configuration
42 dipoles inside, 18 dipoles outside brain

I get the error

??? Reference to non-existent field 'cumtapcnt'.

Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
  Ntrials = length(freq.cumtapcnt);

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);

when i use the mtmfft i get the cumtapcnt values
i get a error when using the mtmfft for source analysis
cfg.frequency=5;
cfg.refchannel='C1';
cfg.refdip=10;
cfg.method='dics';
cfg.grid=grid;
cfg.elec=elec1;
cfg.vol=vol;
[source]=sourceanalysis(cfg,freqmtmfft);
using headmodel specified in the configuration
using electrodes specified in the configuration
610 dipoles inside, 600 dipoles outside brain

??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
The cross-spectral-density matrix is not complete

Error in ==> sourceanalysis at 663
    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);


Please go through it
Because for the DICS i would like to use the freqmtmwelch
is this a problem with freqmtmwelch
or something wrong in my inputs

thanking you

with regards
muthuraman


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:18:48 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:18:48 +0200
Subject: DICS
In-Reply-To: <BAY121-F235CB8F83C35EABC33B41ECE620@phx.gbl>
Message-ID: <TUE.25.JUL.2006.121848.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:29:23 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:29:23 +0200
Subject: sourcestatistics question
In-Reply-To: <44C5D532.4020708@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.122923.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris,

On 25 Jul 2006, at 10:24, Floris de Lange wrote:
> I have a question concerning sourcestatistics.
> I have calculated the source localization using DICS for 11
> individual subjects. Now I would like to test for the significance
> of the sources using sourcestatistics. The examples I find in the
> tutorial are about checking for significance within one subject by
> shuffling the individual trials. However, what I would like to do
> is to do the statistics on the 2nd level (subject), and to test for
> the consistency of the source reconstructions across subjects.
> Is this possible with sourcestatistics?

Yes, that is possible.

> If so, how should I feed the data to sourcestatistics? If I say
> something like stats = sourcestatistics(cfg,sources{1:11}) where
> each source{} is a normalized source (calculated by sourceanalysis,
> normalized with sourceinterpolate), it complains (rightfully) that
> there's a reference to the non-existing field 'statistic'.


Assuming you want to use a randomization test to get a Monte-Carlo
estimate of the probability of your null-hypothesis
(cfg.method='montecarlo'), cfg.statistic should be a string
describing the statistic whose randomization distribution you want to
compute. See the help of statistic_montecarlo.

Furthermore, the hypothesis that you want to test is about two
conditions. That means that the two conditions both should be fed
into sourcestatistics, e.g. something like

for i=1:11
   read data from condition 1 for each subject
   cond1{i} = sourceanalysis(...)
   cond1{i} = sourceinterpolate(...)
   cond1{i} = volumenormalize(...)
   read data from condition 2 for each subject
   cond2{i} = sourceanalysis(...)
   cond2{i} = sourceinterpolate(...)
   cond2{i} = volumenormalize(...)
end

cfg = []
cfg.design = [1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2    %
condition number
               1 2 3 4 5 .....       1 2 3 4 5 ....       ]   %
subject number
cfg.method = 'montecarlo'
cfg.statistic = 'depsamplesT'
cfg.ivar = 1; % row with the independent variable (condition) in the
design
cfg.uvar = 2; % row with the unit of observation (subject) in the
design matrix
stat = sourcestatistics(cfg, cond1{:}, cond2{:});

It is also possible to do other tests, which are more in line with a
2nd level analysis (e.g. compute t-maps per subject and test over
subjects). It is better to discuss those face-to-face.

best regards,
Robert


From Jan.Schoffelen at FCDONDERS.RU.NL  Tue Jul 25 12:32:38 2006
From: Jan.Schoffelen at FCDONDERS.RU.NL (Jan Mathijs Schoffelen)
Date: Tue, 25 Jul 2006 12:32:38 +0200
Subject: DICS
In-Reply-To: <4442BD57-7F97-469F-990B-CD63E171E371@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.123238.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Muthuraman,

One little addition to the freq.cumtapcnt-story.
freq.cumtapcnt should be a vector instead of a scalar:
the correct way to specify it would be: freq.cumtapcnt = ones(ntrl,1)
with ntrl the number of trials in your data.

Yours,

Jan-Mathijs

-----Original Message-----
From: FieldTrip discussion list [mailto:FIELDTRIP at NIC.SURFNET.NL] On Behalf
Of Robert Oostenveld
Sent: Tuesday, July 25, 2006 12:19 PM
To: FIELDTRIP at NIC.SURFNET.NL
Subject: Re: [FIELDTRIP] DICS

Hi Muthuraman

The cumulative number of tapers per trial is indeed not kept in
mtmwelch. You should specify it manually prior to freqanalysis. I do
not recall whether you use keeptrials and keeptapers but if not, then
it is safe to specify freq.cumtapcnt=1.

Keeping count of the number of trials and tapers in the various
freqanalysis methods is not done consistenly, it is something that we
will improve in the future.

best regards,
Robert


On 17 Jul 2006, at 16:12, Muthuraman Muthuraman wrote:

> I get the error
>
> ??? Reference to non-existent field 'cumtapcnt'.
>
> Error in ==> fieldtrip-20060712\private\prepare_freq_matrices at 124
>  Ntrials = length(freq.cumtapcnt);
>
> Error in ==> sourceanalysis at 663
>    [Cf, Cr, Pr, Ntrials, cfg] = prepare_freq_matrices(cfg, data);
>
> when i use the mtmfft i get the cumtapcnt values
> but not when using the mtmwelch is this a problem in the mtmwelch
> or i am doing something wrong


From r.oostenveld at FCDONDERS.RU.NL  Tue Jul 25 12:32:52 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Tue, 25 Jul 2006 12:32:52 +0200
Subject: Source anaylsis
In-Reply-To: <BAY121-F324EA9761CB16729FB47E8CE5A0@phx.gbl>
Message-ID: <TUE.25.JUL.2006.123252.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:

> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
> The cross-spectral-density matrix is not complete

this indicates that you did not include all channel combinations
(cfg.channelcmb) in freqanalysis, and that you try to use a cross
spectral density matrix for DISC that is not complete (i.e. come
cross-spectra are missing). You should correct your configuration for
the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
CHANNELCOMBINATION function).

best regards,
Robert


From jciveira at UNAV.ES  Tue Jul 25 14:22:36 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Tue, 25 Jul 2006 14:22:36 +0200
Subject: Source anaylsis
In-Reply-To: <77E33A56-A0D4-43FC-94E0-91A3AC31197B@fcdonders.ru.nl>
Message-ID: <TUE.25.JUL.2006.142236.0200.FIELDTRIP@NIC.SURFNET.NL>

I have the same problem and I do not know how to fix it, even though I  
tried your solution with cfg.channelcmb.

My case is : i have a set of 82 electrodes, I make freqanalysis with  
cfg.channelcmb = channelcombination({'all' 'all'},data.label);  thus I  
get

freq =

         label: {82x1 cell}
        dimord: 'chan_freq'
          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000 16.8000  
17.2000 17.6000 18]
     powspctrm: [82x11 double]
      labelcmb: {3321x2 cell}
     crsspctrm: [3321x11 double]
           cfg: [1x1 struct]

When I make sourceanalysis with a dipole as a reference and I have no  
problems but when I want to use a channel as a reference I always have  
the same error:
The cross-spectral-density matrix is not complete.

Can you help me?

Best regards
Juan

Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 10:34, Muthuraman Muthuraman wrote:
>
>> ??? Error using ==> fieldtrip-20060712\private\prepare_freq_matrices
>> The cross-spectral-density matrix is not complete
>
> this indicates that you did not include all channel combinations
> (cfg.channelcmb) in freqanalysis, and that you try to use a cross
> spectral density matrix for DISC that is not complete (i.e. come
> cross-spectra are missing). You should correct your configuration for
> the frequencyanalysus (e.g. cfg.channelcmb = {'all' 'all'} or see the
> CHANNELCOMBINATION function).
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From stephan.bickel at ANATOM.UNIZH.CH  Tue Jul 25 17:39:32 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel)
Date: Tue, 25 Jul 2006 17:39:32 +0200
Subject: import bva time-freq data
Message-ID: <TUE.25.JUL.2006.173932.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi,

I try to import brainvision analyzer dat files of wavelet time-frequency
decompositions into fieldtrip. I exported an average file, so it consists
only of a single segment. However I have some problems with it I am not sure
if it is because I define the segment wrong or if it is not possible to import
this data into fieldtrip.
I pasted the code and the error message below and attached the marker- and
header file.

I would be really glad for any suggestions.

Thank you very much,

Stephan 



This is what I tried:

cfg = [];
cfg.headerfile =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vhdr';
cfg.datafile   =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.dat';
cfg.channel = 'all';

%Trialdefinition
cfg.trialdef.trgfile  =...
'c:\HumanEEG_data\Vision_Exports\WT_all_ft_test.vmrk';
cfg.trialdef.stim  = 'New Segment';
%cfg.trialdef.segment  = 'yes';
cfg.trialdef.timezero = 'yes';
% cfg.trialdef.eventtype  = 'Time 0';
cfg.trialdef.prestim    = 0.500;
cfg.trialdef.poststim   = 1.600;

[cfg] = definetrial(cfg)
[raw] = preprocessing(cfg)


and this is the error message:


evaluating trialfunction 'trialfun_brainvision'
0 stimulus markers converted into 0 trials
0 response markers converted into 0 trials
1 segment markers converted into 0 trials
1 "Time 0" markers converted into 1 trials
found 0 events
created 1 trials

cfg = 

    headerfile: [1x88 char]
      datafile: [1x87 char]
       channel: 'all'
      trialdef: [1x1 struct]
       dataset: []
      trialfun: 'trialfun_brainvision'
         event: []
           trl: [1 1051 -250]
       version: [1x1 struct]

retaining exist trial definition
retaining exist event information
found 0 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060330\private\read_brainvision_eeg at 65
    dat(line,:) = str2num(str);

Error in ==> read_fcdc_data at 419
  dat = read_brainvision_eeg(datafile, hdr, begsample, endsample);

Error in ==> preprocessing at 322
  dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);

-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vmrk
Type: application/octet-stream
Size: 741 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment-0004.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: WT_all_ft_test.vhdr
Type: application/octet-stream
Size: 3459 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20060725/20c93631/attachment-0005.obj>

From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:07:03 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:07:03 +0200
Subject: Source anaylsis
In-Reply-To: <20060725142236.lz6o47hwkkwo4sw4@webmail.unav.es>
Message-ID: <WED.26.JUL.2006.090703.0200.FIELDTRIP@NIC.SURFNET.NL>

On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
> I have the same problem and I do not know how to fix it, even
> though I tried your solution with cfg.channelcmb.
>
> My case is : i have a set of 82 electrodes, I make freqanalysis
> with cfg.channelcmb = channelcombination({'all'
> 'all'},data.label);  thus I get
>
> freq =
>
>         label: {82x1 cell}
>        dimord: 'chan_freq'
>          freq: [14 14.4000 14.8000 15.2000 15.6000 16 16.4000
> 16.8000 17.2000 17.6000 18]
>     powspctrm: [82x11 double]
>      labelcmb: {3321x2 cell}
>     crsspctrm: [3321x11 double]
>           cfg: [1x1 struct]
>
> When I make sourceanalysis with a dipole as a reference and I have
> no problems but when I want to use a channel as a reference I
> always have the same error:
> The cross-spectral-density matrix is not complete.
>
> Can you help me?
>
> Best regards
> Juan

Dear Juan

Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
expect the latter, since computing coherence between one particular
EEG electrode and sources in the brain with DICS will not lead to
meaningfull results. The selected EEG electrode will pick up volume-
conducted activity from all sources in the brain (as opposed to an
EMG electrode) and would be referenced to the same reference
electrode as all other electrodes. So there will be a huge amount of
trivial coherence, whose spatial distribution in the brain will
probably more reflect the volume-conduction than something
physiologically interpretable.

Although I have not used DICS with a reference channel on EEG data, I
would expect the code to work if you follow a procedure like this:

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EEG electrodes
cfg.implicitref = name of implicit reference electrode used for the
EEG channels
cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
data_eeg = preprocessing(cfg)

cfg = ... trial selection stuff
cfg = definetrial(cfg)
cfg = ... preprocessing stuff
cfg.channel = list of all EMG electrodes, assuming that they are
bipolar channels
% cfg.reref may be used if you have two unipolar channels on the
muscle of interest
cfg.rectify = 'yes'   % this makes the spiky nature of the EMG more
visible, and improves the coherence estimate
cfg.baseline = 'yes'  % remove baseline prior to rectification, not
neccessary but still preferable
data_emg = preprocessing(cfg)

% combine EEG and EMG in one dataset, both are referenced differently
data = appenddata([], data_eeg, data_emg);

cfg = ... frequency analysis stuff
cfg.ouptu = 'powandcsd'
cfg.channelcmb = {'all', 'all'}
freq = freqanalysis(cfg, data);

cfg = ... source specific settings
cfg.refchan = one of your bipolar EMG channels
cfg.channel = all your EEG channels
source = sourceanalysis(cfg, freq)

best regards,
Robert


From r.oostenveld at FCDONDERS.RU.NL  Wed Jul 26 09:46:59 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Wed, 26 Jul 2006 09:46:59 +0200
Subject: import bva time-freq data
In-Reply-To: <LISTSERV%200607251739327460.3C55@NIC.SURFNET.NL>
Message-ID: <WED.26.JUL.2006.094659.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Stephan,

On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
> I try to import brainvision analyzer dat files of wavelet time-
> frequency
> decompositions into fieldtrip. I exported an average file, so it
> consists
> only of a single segment. However I have some problems with it I am
> not sure
...
> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
> \WT_all_ft_test.vmrk';
> cfg.trialdef.stim  = 'New Segment';
> %cfg.trialdef.segment  = 'yes';
> cfg.trialdef.timezero = 'yes';
> % cfg.trialdef.eventtype  = 'Time 0';

In general I would recommend to use the trialfun_general instead of
the trialfun_brainvision (but the latter should also work). The
trialfun_general was written later and will work for any dataformat
that is supported by read_fcdc_event, the trialfun_brainvision is
only there to support old scripts (i.e. scripts that predate the
read_fcdc_event function).

If not specified manually as trialfun, the trialfun_general or
trialfun_brainvision is selected automatically based on the cfg
settings. In your case you could do
    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
    cfg.trialdef.prestim    = number, latency in seconds (optional)
    cfg.trialdef.poststim   = number, latency in seconds (optional)
See teh help of DEFINETRIAL.

But in your case this is not the problem. The problem lies in
> cfg.trialdef.prestim    = 0.500;
> cfg.trialdef.poststim   = 1.600;

If I do
   event = read_fcdc_event('WT_all_ft_test.vmrk')
then I see that

 >> event(1)
ans =
         type: 'New Segment'
        value: '1'
       sample: 1
     duration: 0
       offset: []

 >> event(2)
ans =
         type: 'Time 0'
        value: '251'
       sample: 1
     duration: 0
       offset: []

The 'New Segment' event is at sample 1, hence you cannot select a
500ms pre-stimulus window before it (you cannot read before the
beginning of the data). The alternative 'Time 0' event is also at
sample 1. It specifies a value of 251, which surprises me. I would
expect that event to be present at sample 251.

Reading the marker file with a text editor, I see (abbreviated) the
explanation
    Mk<number>=<Type>,<Description>,<Position>,<Size>,
<Channelnumber>,<Date>
and the markers
    Mk1=New Segment,,1,1,0,00000000000000000000
    Mk2=Time 0,,251,1,0
So Mk2 should indeed be at sample 251.

With the matlab debugger in the private/read_event.m file (around
line 164) I notice that the problem lies in the missing value of the
second field, i.e.
    Mk2=Time 0,,251,1,0
is interpreted as
    Mk2=Time 0,251,1,0  (one comma less)
All fields are shifted by one, resulting in the sample number being
interpreted as the description (i.e. value).

I have fixed it in read_event, by replacing line 151 from
   tok = tokenize(line, '=');
into
   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
seperator
and by replacing line 157 from
   tok = tokenize(tok{2}, ',');
into
   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
the seperator
i.e. both calls to the tokenize function now use the third argument.
You can apply the same change to your copy, but please check that you
have an up to date version of tokenize that accepts 3 input arguments
(tokenize is in private as well). I will include the bug fix in the
upcoming nightly release of FieldTrip. Better upgrade tomorrow to the
latest FT version from the FTP server.

I think that sofar it was not noticed since people here at the
Donders typically use triggers as markers in their Brainvision data,
and in case of triggers the second field would not be empty.

If I now do
   cfg = [];
   cfg.dataset = 'WT_all_ft_test.vhdr'
   cfg.trialdef.eventtype = 'Time 0'
   cfg.trialdef.prestim  =  0.5000
   cfg.trialdef.poststim =  1
   cfg = definetrial(cfg)
I get
   evaluating trialfunction 'trialfun_general'
   found 2 events
   created 1 trials
   cfg =
        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: []
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
which is correct. The trial (cfg.trl) runs from sample 1 to sample
750 in the data file, and the first sample of that trial corresponds
with time cfg.trl(1,3)/fsample=-0.500 seconds.

Thanks for reporting the bug,
Robert


From stephan.bickel at ANATOM.UNIZH.CH  Wed Jul 26 17:28:54 2006
From: stephan.bickel at ANATOM.UNIZH.CH (Stephan Bickel, Anatomisches Inst.)
Date: Wed, 26 Jul 2006 11:28:54 -0400
Subject: import bva time-freq data
In-Reply-To: <30468A0B-0B5C-41EF-9FB3-84FD50036A22@fcdonders.ru.nl>
Message-ID: <WED.26.JUL.2006.112854.0400.FIELDTRIP@NIC.SURFNET.NL>

Hi Robert,

thank you very much for your response and for fixing the bug. Reading in the
trial information works fine now. Unfortunately I encounter another problem
when I want to read in the data with preprocessing. I wonder if it is yet
possible to read in time-frequency data from bva-dat files?

Sorry for bothering you again and thank you very much in advance.

Stephan


So when I do:

cfg = [];
cfg.dataset = 'WT_all_ft_test.vhdr'
cfg.trialdef.eventtype = 'Time 0'
cfg.trialdef.prestim = 0.5000
cfg.trialdef.poststim = 1
cfg = definetrial(cfg)
cfg.datafile   = 'WT_all_ft_test.dat';
cfg.channel = 'all';
[raw] = preprocessing(cfg)


I get:

cfg =

        dataset: 'WT_all_ft_test.vhdr'
       trialdef: [1x1 struct]
       datafile: 'WT_all_ft_test.dat'
     headerfile: 'WT_all_ft_test.vhdr'
       trialfun: 'trialfun_general'
          event: [1x2 struct]
            trl: [1 750 -250]
        version: [1x1 struct]
        channel: 'all'

retaining exist trial definition
retaining exist event information
found 2 events
created 1 trials
rejected    0 trials completely
rejected    0 trials partially
resulting   1 trials
reading and preprocessing
reading and preprocessing trial 1 from 1
??? Subscripted assignment dimension mismatch.

Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
     dat(line,:) = str2num(str);

Error in ==> fieldtrip-20060725\private\read_data at 239
     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Error in ==> read_fcdc_data at 49
[dat] = read_data(varargin{:});

Error in ==> preprocessing at 366
   dat = read_fcdc_data(cfg.datafile, hdr, begsample, endsample, rawindx,
iscontinuous);



On Wed, 26 Jul 2006 09:46:59 +0200
  Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> wrote:
> Hi Stephan,
>
> On 25 Jul 2006, at 17:39, Stephan Bickel wrote:
>> I try to import brainvision analyzer dat files of wavelet time-
>> frequency
>> decompositions into fieldtrip. I exported an average file, so it
>> consists
>> only of a single segment. However I have some problems with it I am
>> not sure
> ...
>> cfg.trialdef.trgfile  = 'c:\HumanEEG_data\Vision_Exports
>> \WT_all_ft_test.vmrk';
>> cfg.trialdef.stim  = 'New Segment';
>> %cfg.trialdef.segment  = 'yes';
>> cfg.trialdef.timezero = 'yes';
>> % cfg.trialdef.eventtype  = 'Time 0';
>
> In general I would recommend to use the trialfun_general instead of  the
>trialfun_brainvision (but the latter should also work). The
> trialfun_general was written later and will work for any dataformat  that
>is supported by read_fcdc_event, the trialfun_brainvision is  only there to
>support old scripts (i.e. scripts that predate the  read_fcdc_event
>function).
>
> If not specified manually as trialfun, the trialfun_general or
> trialfun_brainvision is selected automatically based on the cfg  settings.
>In your case you could do
>    cfg.trialdef.eventtype  = 'New Segment' or 'Time 0'
>    cfg.trialdef.prestim    = number, latency in seconds (optional)
>    cfg.trialdef.poststim   = number, latency in seconds (optional)
> See teh help of DEFINETRIAL.
>
> But in your case this is not the problem. The problem lies in
>> cfg.trialdef.prestim    = 0.500;
>> cfg.trialdef.poststim   = 1.600;
>
> If I do
>   event = read_fcdc_event('WT_all_ft_test.vmrk')
> then I see that
>
> >> event(1)
> ans =
>         type: 'New Segment'
>        value: '1'
>       sample: 1
>     duration: 0
>       offset: []
>
> >> event(2)
> ans =
>         type: 'Time 0'
>        value: '251'
>       sample: 1
>     duration: 0
>       offset: []
>
> The 'New Segment' event is at sample 1, hence you cannot select a  500ms
>pre-stimulus window before it (you cannot read before the  beginning of the
>data). The alternative 'Time 0' event is also at  sample 1. It specifies a
>value of 251, which surprises me. I would  expect that event to be present
>at sample 251.
>
> Reading the marker file with a text editor, I see (abbreviated) the
> explanation
>    Mk<number>=<Type>,<Description>,<Position>,<Size>,
> <Channelnumber>,<Date>
> and the markers
>    Mk1=New Segment,,1,1,0,00000000000000000000
>    Mk2=Time 0,,251,1,0
> So Mk2 should indeed be at sample 251.
>
> With the matlab debugger in the private/read_event.m file (around  line
>164) I notice that the problem lies in the missing value of the  second
>field, i.e.
>    Mk2=Time 0,,251,1,0
> is interpreted as
>    Mk2=Time 0,251,1,0  (one comma less)
> All fields are shifted by one, resulting in the sample number being
> interpreted as the description (i.e. value).
>
> I have fixed it in read_event, by replacing line 151 from
>   tok = tokenize(line, '=');
> into
>   tok = tokenize(line, '=', 0);  % do not squeeze repetitions of the
> seperator
> and by replacing line 157 from
>   tok = tokenize(tok{2}, ',');
> into
>   tok = tokenize(tok{2}, ',', 0);    % do not squeeze repetitions of
> the seperator
> i.e. both calls to the tokenize function now use the third argument.  You
>can apply the same change to your copy, but please check that you  have an
>up to date version of tokenize that accepts 3 input arguments  (tokenize is
>in private as well). I will include the bug fix in the  upcoming nightly
>release of FieldTrip. Better upgrade tomorrow to the  latest FT version
>from the FTP server.
>
> I think that sofar it was not noticed since people here at the  Donders
>typically use triggers as markers in their Brainvision data,  and in case
>of triggers the second field would not be empty.
>
> If I now do
>   cfg = [];
>   cfg.dataset = 'WT_all_ft_test.vhdr'
>   cfg.trialdef.eventtype = 'Time 0'
>   cfg.trialdef.prestim  =  0.5000
>   cfg.trialdef.poststim =  1
>   cfg = definetrial(cfg)
> I get
>   evaluating trialfunction 'trialfun_general'
>   found 2 events
>   created 1 trials
>   cfg =
>        dataset: 'WT_all_ft_test.vhdr'
>       trialdef: [1x1 struct]
>       datafile: []
>     headerfile: 'WT_all_ft_test.vhdr'
>       trialfun: 'trialfun_general'
>          event: [1x2 struct]
>            trl: [1 750 -250]
>        version: [1x1 struct]
> which is correct. The trial (cfg.trl) runs from sample 1 to sample  750 in
>the data file, and the first sample of that trial corresponds  with time
>cfg.trl(1,3)/fsample=-0.500 seconds.
>
> Thanks for reporting the bug,
> Robert


From r.oostenveld at FCDONDERS.RU.NL  Thu Jul 27 09:26:26 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Thu, 27 Jul 2006 09:26:26 +0200
Subject: import bva time-freq data
In-Reply-To: <web-4062129@idmailbe2.unizh.ch>
Message-ID: <THU.27.JUL.2006.092626.0200.FIELDTRIP@NIC.SURFNET.NL>

On 26 Jul 2006, at 17:28, Stephan Bickel, Anatomisches Inst. wrote:
> thank you very much for your response and for fixing the bug.
> Reading in the trial information works fine now. Unfortunately I
> encounter another problem when I want to read in the data with
> preprocessing. I wonder if it is yet possible to read in time-
> frequency data from bva-dat files?

...

> Error in ==> fieldtrip-20060725\private\read_brainvision_eeg at 65
>     dat(line,:) = str2num(str);
>
> Error in ==> fieldtrip-20060725\private\read_data at 239
>     dat = read_brainvision_eeg(filename, hdr, begsample, endsample);

Hi Stephan,

I have no idea how time-frequency data is stored in the BVA *.dat
files. Even if the low-level reading function were to support it, the
PREPROCESSING function at least would not be able to deal with it,
since the data structure returned by preprocessing does not allow for
3D data (channelXtimeXfrequency). That type of data would have to be
stored in a structure that is compatible with the output of
FREQANALYSIS. The preprocessing function is made for raw continuous
(or trial-based) data, and it happens to work as well for averaged
ERP data, although it is not specifically made for that.

For other external software (BESA, EEGLAB, LORETA) I have created
xxx2fieldtrip functions, i.e. functions that import already processed
data and format it into the fieldtrip structure that is most
compatible with it. In this case I can imagine a
BRAINVISION2FIELDTRIP function that would read your dat file and
return a "freq" structure. Besides making a brainvision2fieldtrip
function, it requires making either the read_brainvision_eeg smarter
(or probably making a seperate read_brainvision_tfr function) and
improving the detection of teh content of the BVA *.dat file. Please
have a look in besa2fieldtrip to see how I approach it there.

best regards,
Robert


From jciveira at UNAV.ES  Thu Jul 27 18:54:27 2006
From: jciveira at UNAV.ES (Juan Civeira)
Date: Thu, 27 Jul 2006 18:54:27 +0200
Subject: Source anaylsis
In-Reply-To: <9A0AA0F8-A8AD-4958-B224-4E3547238053@FCDONDERS.RU.NL>
Message-ID: <THU.27.JUL.2006.185427.0200.FIELDTRIP@NIC.SURFNET.NL>

   Dear Robert:

   Thanks for your answer but my problems are still there. The main
problem is that i can not use DICS with a reference channel on EEG
data, no matter if  the reference is an EEG channel or EMG or some
other channel.

   I have taken my measures with 85 electrodes, some of them are
EEG1010, some EOG and some other that I have placed myself.

   I do the freq analysis and then the source analysis. Then I can
make another source analysis with refdip but there is always errors
when I use refchan. I do not think is a problem of the reference ,
cos always the program says that the  cross-spectral-density with
the reference channel is not complete .

   I have done the freq analysis making cfg.channelcmb combinations of
all channels with only one, so i make sure that the
cross-spectral-density is complete,but when I make source analysis
with that channel as a reference is the same.

   Searching in the code I have come up with the fact that the error
is always the condition in line 195 of the function
prepare_freq_matrices (version 2006/03/08). My idea is that there is
a confussion because if we have a set of N channels and we make  
combinations with channelcombination, the result is N-1 combinations  
because we do not include the combination of the reference with itself.

I hope i explain myself clear enough, anyway I am trying to fix the  
problem by myself but any help or suggestion will be great

Juan


Robert Oostenveld <r.oostenveld at FCDONDERS.RU.NL> ha escrito:

> On 25 Jul 2006, at 14:22, jciveira at unav.es wrote:
>> I have the same problem and I do not know how to fix it, even   
>> though I tried your solution with cfg.channelcmb.
>>
>> My case is : i have a set of 82 electrodes, I make freqanalysis   
>> with cfg.channelcmb = channelcombination({'all'
'all'},data.label);
>>  thus I get
>>
>> freq =
>>
>>        label: {82x1 cell}
>>       dimord: 'chan_freq'
>>         freq: [14 14.4000 14.8000 15.2000 15.6000 16
16.4000
>> 16.8000 17.2000 17.6000 18]
>>    powspctrm: [82x11 double]
>>     labelcmb: {3321x2 cell}
>>    crsspctrm: [3321x11 double]
>>          cfg: [1x1 struct]
>>
>> When I make sourceanalysis with a dipole as a reference and I
have
>> no problems but when I want to use a channel as a reference I   
>> always have the same error:
>> The cross-spectral-density matrix is not complete.
>>
>> Can you help me?
>>
>> Best regards
>> Juan
>
> Dear Juan
>
> Do you have 82 EEG electrodes, or 81 and one EMG electrode? I would
> expect the latter, since computing coherence between one particular
EEG
> electrode and sources in the brain with DICS will not lead to
> meaningfull results. The selected EEG electrode will pick up
> volume-conducted activity from all sources in the brain (as opposed
to
> an EMG electrode) and would be referenced to the same reference
> electrode as all other electrodes. So there will be a huge amount
of
> trivial coherence, whose spatial distribution in the brain will
> probably more reflect the volume-conduction than something
> physiologically interpretable.
>
> Although I have not used DICS with a reference channel on EEG data,
I
> would expect the code to work if you follow a procedure like this:
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EEG electrodes
> cfg.implicitref = name of implicit reference electrode used for the
EEG
> channels
> cfg.reref = 'all'  % rereference all EEG electrodes to an average
reference
> data_eeg = preprocessing(cfg)
>
> cfg = ... trial selection stuff
> cfg = definetrial(cfg)
> cfg = ... preprocessing stuff
> cfg.channel = list of all EMG electrodes, assuming that they are
> bipolar channels
> % cfg.reref may be used if you have two unipolar channels on the
muscle
> of interest
> cfg.rectify = 'yes'   % this makes the spiky nature of the EMG
more
> visible, and improves the coherence estimate
> cfg.baseline = 'yes'  % remove baseline prior to rectification,
not
> neccessary but still preferable
> data_emg = preprocessing(cfg)
>
> % combine EEG and EMG in one dataset, both are referenced
differently
> data = appenddata([], data_eeg, data_emg);
>
> cfg = ... frequency analysis stuff
> cfg.ouptu = 'powandcsd'
> cfg.channelcmb = {'all', 'all'}
> freq = freqanalysis(cfg, data);
>
> cfg = ... source specific settings
> cfg.refchan = one of your bipolar EMG channels
> cfg.channel = all your EEG channels
> source = sourceanalysis(cfg, freq)
>
> best regards,
> Robert



----------------------------------------------------------------
Este mensaje ha sido enviado desde https://webmail.unav.es


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 09:49:55 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 09:49:55 +0200
Subject: Source anaylsis
In-Reply-To: <20060727185427.3r2gdy6zkk0o4og8@webmail.unav.es>
Message-ID: <FRI.28.JUL.2006.094955.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Juan

>   I do the freq analysis and then the source analysis. Then I can
> make another source analysis with refdip but there is always errors
> when I use refchan. I do not think is a problem of the reference ,
> cos always the program says that the  cross-spectral-density with
> the reference channel is not complete .
>
>   I have done the freq analysis making cfg.channelcmb combinations of
> all channels with only one, so i make sure that the
> cross-spectral-density is complete,but when I make source analysis
> with that channel as a reference is the same.
>
>   Searching in the code I have come up with the fact that the error
> is always the condition in line 195 of the function
> prepare_freq_matrices (version 2006/03/08). My idea is that there is
> a confussion because if we have a set of N channels and we make
> combinations with channelcombination, the result is N-1
> combinations because we do not include the combination of the
> reference with itself.

Say for simplicity that you want EMG as refchan. In freqanalysis,
cfg.channelcmb should be all with all. You need the CSD between all
EEG channels (all EEG with all EEG), and of all EEG channels with the
EMG. In prepare_freq_matrices the CSD matrix (Cf) of EEG with EEG is
made (Neeg X Neeg), the CSD vector (Cr) of all EEG channels with the
reference EMG channel (Neeg X 1), and the power of the reference
channel (Pr). Together they could also be shaped in a (Neeg+1) X (Neeg
+1) matrix in principle, but that is not done for DICS.

I have tried it out myself, and cannot reproduce your problem. Please
find my test script below. It uses some simulated EEG data. The
beginning of the script is copied from http://www2.ru.nl/fcdonders/
fieldtrip/doku.php?
id=fieldtrip:documentation:compute_forward_simulated_data_and_apply_a_di
pole_fit

The initial part generates raw EEG data, then I add a EMG channel
(with noise, but it is only to demonstrate that fieldtrip does not
give an error), then I do freqanalysis and sourceanalysis. Please try
out and compare my cfgs with yours, since I suspect your
configuration to be the problem.

best regards,
Robert


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% create a set of electrodes, randomly placed on the sphere
elec = [];
elec.pnt = randn(128,3);
dum = sqrt(sum(elec.pnt.^2,2));
elec.pnt = elec.pnt ./ [dum dum dum];  % scale them to a unit sphere
for i=1:128
    elec.label{i} = sprintf('%03d', i);
end

% create a concentric 3-sphere volume conductor, the radius is the
same as for the electrodes
vol = [];
vol.r = [0.88 0.92 1.00]; % radii of spheres
vol.c = [1 1/80 1];       % conductivity
vol.o = [0 0 0];          % center of sphere

% create a dipole simulation with one dipole and a 10Hz sine wave
cfg      = [];
cfg.vol  = vol;             % see above
cfg.elec = elec;            % see above
cfg.dip.pos = [0 0.5 0.3];
cfg.dip.mom = [1 0 0]';     % note, it should be transposed
cfg.dip.frequency = 10;
cfg.ntrials = 10;
cfg.triallength = 1;        % seconds
cfg.fsample = 250;          % Hz
raw1 = dipolesimulation(cfg);

using headmodel specified in the configuration
using electrodes specified in the configuration
selected 128 electrodes
computing simulated data
RMS value of simulated data is 0.0746553

%%%%%%%%%%%%%%%%%%%%%%%%
%%%%% Here it starts deviating from the demo script on the wiki %%%%
%%%%%%%%%%%%%%%%%%%%%%%%

% add a fake EMG channel, containing noise
for i=1:10
   raw1.trial{i}(end+1,:) = randn(1,250);
end
raw1.label{end+1} = 'EMG'

raw1 =

       trial: {1x10 cell}
        time: {1x10 cell}
        elec: [1x1 struct]
     fsample: 250
       label: {1x129 cell}
         cfg: [1x1 struct]

cfg = []
cfg.method = 'mtmfft';
cfg.taper = 'hanning';
cfg.foilim = [2 20];
cfg.output = 'powandcsd';
cfg.channelcmb = {'all', 'all'}

cfg =

         method: 'mtmfft'
          taper: 'hanning'
         foilim: [2 20]
         output: 'powandcsd'
     channelcmb: {'all'  'all'}

freq1 = freqanalysis(cfg, raw1)
WARNING: using only one taper for specified smoothing
processing trial 1, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 2, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 3, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 4, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 5, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 6, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 7, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 8, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 9, nfft: 250 samples, taper length: 250 samples, 1
tapers
processing trial 10, nfft: 250 samples, taper length: 250 samples, 1
tapers

freq1 =
          label: {1x129 cell}
        dimord: 'chan_freq'
          freq: [2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20]
     powspctrm: [129x19 double]
      labelcmb: {8256x2 cell}
     crsspctrm: [8256x19 double]
          elec: [1x1 struct]
           cfg: [1x1 struct]

cfg = [];
cfg.vol = vol;
cfg.freq = 10;
cfg.frequency = 10;
cfg.method = 'dics';
cfg.refchan = 'EMG';
cfg.grid.resolution = 0.3

cfg =
           vol: [1x1 struct]
          freq: 10
     frequency: 10
        method: 'dics'
       refchan: 'EMG'
          grid: [1x1 struct]

source1 = sourceanalysis(cfg, freq1)
using headmodel specified in the configuration
using electrodes specified in the data
selected 128 electrodes
106 dipoles inside, 237 dipoles outside brain
making tight grid
106 dipoles inside, 110 dipoles outside brain
scanning repetition 1
scanning grid
Warning: cross-spectral density matrix is rank deficient
 > In <a href="error:/Users/roberto/matlab/forwinv/beamformer.m,
270,1">beamformer at 270</a>
   In <a href="error:/Users/roberto/matlab/fieldtrip/sourceanalysis.m,
806,1">sourceanalysis at 806</a>
total time in sourceanalysis 75.4 seconds

source1 =
         xgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         ygrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
         zgrid: [-0.7000 -0.4000 -0.1000 0.2000 0.5000 0.8000]
           dim: [6 6 6]
           vol: [1x1 struct]
     frequency: 10
           pos: [216x3 double]
        inside: [1x106 double]
       outside: [1x110 double]
        method: 'average'
           avg: [1x1 struct]
           cfg: [1x1 struct]

diary off


From floris.delange at FCDONDERS.RU.NL  Fri Jul 28 13:01:00 2006
From: floris.delange at FCDONDERS.RU.NL (Floris de Lange)
Date: Fri, 28 Jul 2006 13:01:00 +0200
Subject: sourcestatistics: what to compare?
Message-ID: <FRI.28.JUL.2006.130100.0200.FIELDTRIP@NIC.SURFNET.NL>

Dear Fieldtrippers,

I have a question about what conditions/comparisons make most sense to
statistically compare with sourcestatistics.
I have 2 conditions, A and B, and each condtion has its own pre-stimulus
baseline: basA and basB.

Now, first I'd like to see which regions are showing stronger beta-band
power during A than the baseline. So I want to compare A to basA.
There's 2 ways of going about it:
1) compare (the grand average of) A.avg.pow with basA.avg.pow
2) compare (the grand average of) the neural activity index (NAI) of A
with the baseline

Any ideas which would be best? Or is this just an empirical issue?
(i.e., take the best)

To compare A and B, one has even more options:
1) compare A.avg.pow with B.avg.pow
2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
3) compare NAI(A) with NAI(B)
4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)

Is there's any principled reason for a priori picking any of the above
choices?

Your input is greatly appreciated,
Kind regards,

Floris


From r.oostenveld at FCDONDERS.RU.NL  Fri Jul 28 16:17:24 2006
From: r.oostenveld at FCDONDERS.RU.NL (Robert Oostenveld)
Date: Fri, 28 Jul 2006 16:17:24 +0200
Subject: sourcestatistics: what to compare?
In-Reply-To: <44C9EE6C.8030003@fcdonders.ru.nl>
Message-ID: <FRI.28.JUL.2006.161724.0200.FIELDTRIP@NIC.SURFNET.NL>

Hi Floris

On 28 Jul 2006, at 13:01, Floris de Lange wrote:
> I have a question about what conditions/comparisons make most sense
> to statistically compare with sourcestatistics.
> I have 2 conditions, A and B, and each condtion has its own pre-
> stimulus baseline: basA and basB.
>
> Now, first I'd like to see which regions are showing stronger beta-
> band power during A than the baseline. So I want to compare A to
> basA. There's 2 ways of going about it:
> 1) compare (the grand average of) A.avg.pow with basA.avg.pow
> 2) compare (the grand average of) the neural activity index (NAI)
> of A with the baseline
>
> Any ideas which would be best? Or is this just an empirical issue?
> (i.e., take the best)

Both are valid, you could take the best. There are some exceptions/
predictions though that I can share with you: the NAI is power
divided by estimated (projected) noise. It is inherently difficult to
estimate the noise level (the default is the smallest singular value
of the CSD matrix, but you can manipulate it with cfg.lambda in
sourceanalysis). If you have different numbers of trials in active
v.s. baseline, the noise estimate probably will be different. So then
you would have NAIa = Pa/Na and NAIb = Pb/Nb, if Pa==Pb and Na~=Nb,
these will be trivially different, which is not what you want
(remember Pa==Pb was assumed).

You can compare the projected noise in both conditions, the source-
noise-estimate is uniformely scaled (i.e. on all voxels the same)
with the estimated noise from the CSD. There is no reason to assume
that the noise is different, so if the noise is different, you would
have to explain that.

> To compare A and B, one has even more options:
> 1) compare A.avg.pow with B.avg.pow
> 2) compare (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
> 3) compare NAI(A) with NAI(B)
> 4) compare NAI(A)/NAI(basA) with NAI(B)/NAI(basB)
>
> Is there's any principled reason for a priori picking any of the
> above choices?

I presume that by "compare" you mean "look at the the difference".

In general you should only use the NAI for plotting data if you only
have one condition (and no baseline). The noise estimate that goes
into the NAI is too poor to be of real value. Although sometines it
looks nice in figures, the NAI tends not to be very robust (it
contains global flucuations that mess up the statistic, although the
figure still looks nice).

The baseline is only interesting if you expect something differently
to happen in the two baselines. If that is the case, you better try
to get another baseline in your experimental design. If you would
combine actA, basA, actB, basB into a single number, and that number
would be significant, you still would not be able to tell whether the
difference is betwen the active conditions, theis baselines, or some
weird combination of the 4 "conditions" that you combine.

Typically there is no reason to assume taht the baselines are
different. So
   (A.avg.pow/basA.avg.pow) with (B.avg.pow/basB.avg.pow)
= (A.avg.pow/basC.avg.pow) with (B.avg.pow/basC.avg.pow), where C
means common
= (A.avg.pow-B.avg.pow)/basC.avg.pow.
I.e. you are comparing A and B directly, with a spatially normalizing
for the different baseline-estimated noise at different source
locations.

Whether you want to spatially normalize (divide by some baseline
estimated-source-power) depends on the leadfields (it is neccessary
if cfg.normalize=no, if you want to do multiple comparison
correction, and if you want to look at an additive effect). If you
want to look at a multiplicative effect, you would look at the ratio
between two conditions instead of the difference. The ratio of the
conditions is best implemented with a log transform, after which you
again can look at the difference between the log-transformed power
estimates in both conditions (log(A/B) = log(A) - log(B)). In the
ratio or log-ratio between the powers in both conditions you have
also corrected for the spatial (depth) noise-bias.

good luck with the stats,
Robert

PS as I am on holiday the next 2 weeks, you could ask Ingrid about
more details on this. She is more or less at the same stage in her
analysis (about to do group sourcestatistics), but I suspect that she
already has more "feel" for these issues. Or ask JM.