[FieldTrip] Smoothing before permutation test

[David Groppe]

Smoothing a time-frequency matrix is just as valid.
I would apply the smoothing to whatever it is you are plugging into the
permutation test as an independent observation (in your case it sounds like
trial averages).
   -D



On Wed, Nov 2, 2016 at 1:02 PM, Ramirez Pasos, <Ramirez_U at ukw.de> wrote:

> Thank you David for your response! Just to be clear: by "smoothing" I'm
> not referring to the smoothing performed when using the
> ft_frequencyanalysis multitaper method, but some kind of spatial smoothing
> on the time-frequency matrix obtained via ft_frequencyanalysis - so that
> according to your suggestion, I would smooth each trial, then average
> trials for each subject/condition, and finally use these for statistical
> evaluation?  Is there a reason why smoothing each trial might be preferable
> to smoothing each subject's trial average?
>
> Thank you so much in advance,
> Uri
> ________________________________
> Von: fieldtrip-bounces at science.ru.nl [fieldtrip-bounces at science.ru.nl]"
> im Auftrag von "David Groppe [david.m.groppe at gmail.com]
> Gesendet: Mittwoch, 2. November 2016 14:43
> An: FieldTrip discussion list
> Betreff: Re: [FieldTrip] Smoothing before permutation test
>
> Since permutation tests exploit correlations between variables to increase
> sensitivity, smoothing each trial will increase your sensitivity.
>     cheers,
>         -David
>
> On Wed, Nov 2, 2016 at 6:41 AM, Ramirez Pasos, <Ramirez_U at ukw.de<mailto:
> Ramirez_U at ukw.de>> wrote:
> Dear Fieldtrippers,
>
> I have a set of subcortical LFP signals from 8 patients, which I have
> analyzed with fieldtrip in order to obtain event-related time-frequency
> plots. Unfortunately, I have only three repetitions for each of my 4
> conditions, so there's a lot of noise in the subject-averages of
> time-frequency data. I'm interested in testing baseline vs activation for
> each condition as well as comparisons between my 4 conditions (A1A2, A1B1,
> B1B2, A2B2).
>
> Could anyone tell me their take on smoothing before permuting? Is that a
> valid procedure with such a small sample size? I've searched for literature
> discussing smoothing before permuting (mostly the Holmes papers), where
> much talk of smoothing refers to concepts such as "locally pooled variance"
> and "pseudo t-statistics," but I don't know how this fits with fieldtrip's
> cluster statistics functions.
>
> Any thoughts would be greatly appreciated!
>
> U. Ramirez
> University of Wuerzburg
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20161102/3cf9b77d/attachment-0002.html>