multitapers

[J.M. Schoffelen]

The number of tapers applied to estimate each time-frequency point is
dependent on the product between the smoothing and the width of your window.
In your case the product is equal to 1 for all frequencies, so you only get
one taper.
I think when you take cfg.tapsmofrq = 2./cfg.t_ftimwin you get 3 tapers per
frequency. The quiz now is to come up with the relationship between the
number of tapers and the amount of data in your window in combination with
the smoothing...
Besides, this rounding thing in your t_ftimwin is not necessary, since in
the end, the routine works with sample-numbers and not with seconds, so by
using the round-operator (I guess to avoid rounding-errors) you just
introduce them.

JM


-----Original Message-----
From: FieldTrip discussion list [mailto:FIELDTRIP at NIC.SURFNET.NL] On Behalf
Of Floris de Lange
Sent: Wednesday, February 09, 2005 4:52 PM
To: FIELDTRIP at NIC.SURFNET.NL
Subject: [FIELDTRIP] multitapers

Dear fieldtrippers,

I am experiencing problems when doing a TFR analysis with multitapers.
I have adjusted my t_ftimwin and tapsmofrq to the frequency that is
calculated, as follows:
cfg.t_ftimwin   =
round(TB_TFR_preproc.fsample.*3./cfg.foi)./TB_TFR_preproc.fsample;
cfg.tapsmofrq   =   1./cfg.t_ftimwin;

Since the frequencies I want to look at are between 15 and 50 Hz
(cfg.foi  = 15:50) I get values for the taper time window between 0.2
and 0.06, and for the taper smoothing frequency between 5 and 16.67 Hz.

My signal length is 1.3 sec, which I pad out to 2 sec.
cfg.toi         =   -1:0.02:0.3;
cfg.pad         =   2;

Now, when I do my multitaper analysis, it warns me that there is only 1
taper, for each frequency. I was under the impression that the taper
length is specified by ftimwin, which ranges from 0.2 to 0.06, so I
don't really understand why it only uses one taper.

Can anybody tell me what I'm doing wrong?
Just for completeness, I pasted the settings I use below.

Thanks,
Floris

% with tapers
load TB_TFR_preproc
cfg = [];
cfg.output      =   'pow';
cfg.sgn         =   'all';
cfg.method      =   'mtmconvol';
cfg.foi         =   15:50;

cfg.t_ftimwin   =
round(TB_TFR_preproc.fsample.*3./cfg.foi)./TB_TFR_preproc.fsample;
cfg.tapsmofrq   =   1./cfg.t_ftimwin;
cfg.toi         =   -1:0.02:0.3;
cfg.pad         =   2;
cfg.keeptrials  =   'no';
cfg.keeptapers  =   'no';

TB_mtmconvol=freqanalysis(cfg, TB_TFR_preproc);
save TB_mtmconvol TB_mtmconvol