Hi!
I am running a single subject decoding analysis with leave one out cross-validation in TDT. I am decoding two classes in 7 ROIs. I am obtaining decoding accuracies per ROI and I ran permutation tests afterwards to check whether the accuracy is above chance level. This is single subject data, not second level.
I am running into the following situation:
One of my ROIs shows a decoding accuracy of about 65%, which is significant according to the permutation test. A second ROI shows a decoding accuracy of 65% as well, but here, it is not signifficantly different from chance. If I now test for whether accuracy in one ROI exceeds accuracy in the other ROI, using the differences between decoding accuracies to make another permutation test, the result is non-significant. Although I recognize that this could be, it seems strange since the empirical chance level seems higher in one ROI than the other. I was wondering whether you had any advice on how to deal with that.
For the permutation, I followed the demo in v3.999, which I think are synchronized permutations if I am not mistaken.
Thanks!