Dear all,
I have a question concerning LOOCV. To avoid circularity bias, I would like to apply the LOOCV method to explore whether there are significant correlations between brain and some behavioral data.
I built my ROIs based on the significant functional activations. Given that I have n participants, I run n t-tests every time leaving out a different subject and then used the ROI coordinates from the significant activations yielded by the t-test to extract the ROI betas from the left-out subject.
Now, I’m not sure what the next step is. Should I simply correlate the n ROI values I got from the previous procedure with behavior? And what happens if, say, I don't find an ROI from one of the t-tests?
Best,
Gabriele
|