[Eeglablist] dftfilt3: timesupport variable fixed

[Makoto Miyakoshi]

Dear Ileana,

That's an interesting question, Ileana.

I did not develop this function, but if you see dftfilt3.m line 122, it
seems cycleinc, winsize, and timesupport override each other, which makes
sense because they are all related to determine the window length. The
default newtimef.m input uses cycleinc which is [3 0.5], which means the
window length is determined by the lowest frequency's 3 cycles, not by 7 SD
of the Gaussian.

> Reducing it at 2 or 3 seems to do the trick, but I would like to
understand sides effects (if any).

The principle is the same as Fourier transform--shorter time window means
less frequency resolution. If you compare the output ERSP plots with cycle
option [2 0.5] and [3 0.5], you can see the difference visually.

Makoto



On Sat, Feb 20, 2021 at 8:02 AM Ileana CNR <ileana.pirovano at itb.cnr.it>
wrote:

> Hi all,
>
> I have a question about time-frequency analysis with newtimef() and wavelet
> method.
>
> Wavelets are actually calculated in dftfilt3.m function, where the variable
> "timesupport" seems to be fixed at the value 7.
>
> If I well understand, this parameter influences the wavelet temporal
> length.
>
>
> When recalled from timefreq(),  "timesupport" cannot be changed by user.
>
> I wonder why it was fixed exactly at 7 and how reducing it could affect the
> results.
>
> For example, in my data (1800 ms, sampling rate = 500 Hz), using such a
> large time standard deviation, results in a too large wavelet window and
> half of the epoch time becomes lost.
>
> Reducing it at 2 or 3 seems to do the trick, but I would like to understand
> sides effects (if any).
>
>
>
> Thank you for any clarification!
>
> Ileana
>
>
>
>
>
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>