[Eeglablist] Multiple Comparison Problem

[Makoto Miyakoshi]

Dear Iman,

Sorry for belated response.

> - 120ms baseline

This seems too short. Unless your data are high-pass filtered at 1000/120 s
= 8Hz, this is not recommended.

> Thus, if I do the ANOVA test, I will have the multiple comparison problem
since I will compare every single time point ( time window) in the baseline
with the after-stim period.
Do you have any suggestions for me to deal with this issue?

See Groppe et al. (2010?) Mass univariate... paper. This explains FWER. I
recommend you try GFWER with u==1, and also try larger numbers. I believe
that if you can impose constrain of minimum continuum of 10 points for
example, using u==10 can be justified (at least qualitatively speaking).

Makoto

On Fri, Dec 16, 2016 at 5:58 PM, Iman Mohammad-Rezazadeh <
irezazadeh at ucdavis.edu> wrote:

> Hi EEGLABers.
>
>
> I wonder if anyone could point me to the correct way of doing the
> following analysis:
>
>
> I have N trials of ERP recording with the following conditions:
>
> - 120ms baseline
>
> - 3000ms after stimulus onset.
>
> - and all stim-locked ;
>
>
>  I would like to figure out whether the stimulus modulates the neural
> activation by finding out the time frames which that significant
> modulations happen.
>
>
> Thus, if I do the ANOVA test, I will have the multiple comparison problem since
> I will compare every single time point ( time window) in the baseline with
> the after-stim period.
>
>
> Do you have any suggestions for me to deal with this issue?
>
>
> Best,
>
> Iman
>
>


-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20161229/64987134/attachment.html>