[FieldTrip] unexpected boundary effects using Morlet wavelets

[Michele Scaltritti]

Dears,

I am Michele Scaltritti, and this is my first post on this mailing list. I
work on ERPs in the context of psycholinguistics, and I just recently
started working with Fieldtrip to perform time-frequency analyses.

I am encountering a problem with respect to boundary effects, as these
affects time windows that I didn't expect.

More precisely, I' running my analysis on epochs going from -1.1s before
stimulus onset (0 s), to 4 s afterwards (for a total length of 5.1s). In
order to do that, I am using Morlet wavelets of 7 cycles, with 50
milliseconds time-steps. I am estimating frequencies ranging from 10 to 35
Hz.

At 10Hz frequency, the length of my wavelet should be 7/10/pi, so around
222 ms. As such, I would expect to be able to estimate frequency power at
-800 ms in my epoch (which corresponds to the starting time of what I would
like to be my baseline): in fact, I would expect to be able to estimate
10Hz frequency power as soon as the third time-step (150 ms), corresponding
to -950 in my epochs. However, I don't get any value for 10Hz frequency at
-800 ms.

Clearly, my reasoning is wrong, but I can't figure out what I am missing.
In the following lines, I'll post the code of the configuration I used to
run the frequency decomposition

cfg = []

    cfg.channel    = 'all';

    cfg.method     = 'wavelet';

    cfg.width      = 7;

    cfg.output     = 'pow';

    cfg.foi        = 10:1:35;

    cfg.toi        = -1.1:0.05:4



TF_data = ft_freqanalysis(cfg,data)

Thank you very much for the attention, and please excuse eventual gross
mistakes in my arguments.


michele
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160115/05b1b8bd/attachment-0001.html>