[pymvpa] How you run permutation test for search-light analysis?

[J.A. Etzel]

I've played with versions of your procedure A before (threshold 
individual subject maps, then calculate group-level frequency maps). It 
strikes me as a justifiable method in some cases.

In the datasets I tried the group-level proportion maps were often 
pretty similar to maps from t-tests. Something I found unsatisfactory 
was the flexibility in deciding how to make and combine the binary maps: 
calculate the proportion of subjects in which a searchlights' accuracy 
is above 0.7 (i.e. a fixed threshold)? Or use a statistically-derived 
threshold? Smooth the binary maps before combining? etc.

Your procedure B (bootstrapping the binary maps) is intriguing. One 
thing I like about overlapping binary maps is that it dilutes the effect 
of outliers (t-tests are very susceptible to outliers). Do you think the 
bootstrapping could make outliers a problem again?

Jo


On 3/26/2012 7:41 PM, Francisco Pereira wrote:
> I tend to go for something like one of the two procedures below
> (Procedure A is analytical, Procedure B is nonparametric). I would
> love to hear comments, objections or to be told I'm completely wrong
> and should do otherwise :)
>
> cheers,
> Francisco