missing cumtapcnt field

[Jan Mathijs Schoffelen]

Hi Ingrid,


>Can I for instance just make a vector of ones with length 76 and put that
>in my freq, or will that create problems?

The answer to this question is yes. Indeed, no cumtapcnt is passed back when
using freqanalysis_mtmconvol. Should be fixed, though.
In your case I think it is sufficient just to provide a vector which has a
dimensionality of Nx1, N being the number of epoch. The content of this
vector will not be used later on in this case.
On the other hand, the cumtapcnt-vector, as outputted by for example
freqanalysis_mtmfft contains for each epoch the number of tapers which have
been used for the estimation of the spectra for that particular epoch.
Please note that if you set the option cfg.keeptapers = 'yes', the vector
consists only of ones (or if you use e.g. a single hanning-taper, or such a
low amount of smoothing, that you only need one 'dpss'-taper, in which case
by the way it does not make sense to use the 'dpss'). In any other case, the
cumtapcnt-vector contains a number, which can be different for each epoch
(when you have trials of different lengths). You can easily compute the
number of tapers, this will be: 2xNxK-1, with N being the length of your
data segment (in seconds), and K being the taper-smoothing you specified in
cfg.tapsmofrq.

Yours,

JM