[pymvpa] effect of signal on null distributions

[J.A. Etzel]

Yes, I just divided by a constant for all voxels in class B, so it is a 
multiplicative change in this case. 1000 permutations, 10 
examples/class/run, partitioning on the runs, 2 or 4 runs. Lengthy 
description (and code) is on the blog. I wrote the R code off the text 
description of normal_feature_dataset; I didn't check to see if it is 
precisely the same as implemented in pyMVPA.

Jo



On 2/15/2013 9:55 AM, MS Al-Rawi wrote:
> Thanks to keep this topic up and for the nice reproduction of
> previous results. I guess changing the SNR could be considered as a
> multiplicative parameter, are you using one SNR value for all voxels?
> To get the full picture, how many permutations did you use? And how
> many samples per each run were there?
>
>
> -Rawi
>