[FieldTrip] monte-carlo on LCMV moments
Stenner, Max-Philipp
max-philipp.stenner at med.ovgu.de
Thu Mar 8 13:26:13 CET 2012
Dear fieldtrip users,
in addition to the question on the new implementation of ft_sourcestatistics for sourcegrandaveraged data which I sent yesterday, I have a second question regarding sourcestatistics on grandaveraged data obtained from LCMV beamforming of ERF.
I would like to perfrom cluster corrected monte-carlo statistics on subj_voxel_time dipole moments, comparable (in its dimensions) to the sensor-level ERF cluster permutation test on subj_channel_time data that is described in the respective tutorial and that worked well.
However, ft_sourcegrandaverage doesn't seem to permit "averaging" (with keepindividual = yes) of parameters that have more than one value per gridpoint (i.e. other than pow, nai etc.).
In addition, the get_source_lcmv_mom subfunction in statistics_wrapper requires 4 rather than 3 dimensions.
Is there any way that subj_voxel_time dipole moment data can be used for monte-carlo statistics comparable to the respective sensor-level ERF test?
Thanks very much,
best wishes
Max
(fieldtrip version: fieldtrip-lite-20111130 on a Windows7 PC)
________________________________________
Von: fieldtrip-bounces at donders.ru.nl [fieldtrip-bounces at donders.ru.nl]" im Auftrag von "Stenner, Max-Philipp [max-philipp.stenner at med.ovgu.de]
Gesendet: Mittwoch, 7. März 2012 18:46
Bis: Email discussion list for the FieldTrip project
Betreff: [FieldTrip] ft_sourcestatistics
Dear fieldtrip users,
I am encountering a problem with ft_sourcestatistics when using output from ft_sourcegrandaverage. My parameter of interest is a logratio: log(mean(moments(time of interest))/mean(moments(baseline))) which I simply named 'pow' to be compatible with ft functions. Apart from this, nothing else deviates from how fieldtrip handles the data/nothing else was changed 'manually'.
The problem arises from the fact that the existence of 10 different subjects in the grandaverage (in the usual 1x10 struct field 'trial') isn't represented in the 'powdimord' field of the data after ft_checkdata(varargin{i}, 'datatype', {'source', 'volume'}, 'feedback', 'no', 'inside', 'index', 'sourcerepresentation', 'new'); in the new implementation of ft_sourcestatistics:
varargin{1}
ans =
(before ft_checkdata:)
pos: [8196x3 double]
var: [1x1 struct]
dimord: 'voxel'
trial: [1x10 struct]
inside: [8196x1 double]
outside: [0x1 double]
df: [8196x1 double]
cfg: [1x1 struct]
(after ft_checkdata:)
pos: [8196x3 double]
var: [1x1 struct]
dimord: 'pos'
inside: [8196x1 double]
outside: [0x1 double]
df: [8196x1 double]
cfg: [1x1 struct]
pow: [8196x10 double]
powdimord: 'pos'
This leads to a (wrong) transpose of the design (originally row1 = uvar, row2 = ivar [with two conditions for a depsamples T test]) and ultimately to the message
'The data must contain at least two units (usually subjects).'
Any help would be much appreciated, thanks very much!
Max
(fieldtrip version: fieldtrip-lite-20111130 on a Windows7 PC)
_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
More information about the fieldtrip
mailing list