Hi,
I have individual subject maps with voxel intensities ranging from 0-1.5 that I have until now been feeding into randomise for group stats. Two sample results of these maps show significant results at vox and tfce uncorrected p < .01, with maximum t values (in tstat maps) around 4.
To be consistent with some of our other analyses that have gone through the Feat pipeline the whole way, I am now interested in applying parametric testing to those maps at the group level. I made individual feat directories containing reg directories, a design file and then inputted the individual subject maps as copes/varcopes in the stats directories. Though feat ran, the results were quite dissimilar to randomise results. With very loose thresholding, nothing came up as significant, and upon examination of tstat maps, t values ranged from -.5 to .5--very low.
I was wondering if copes had to have means centered around 0, or if anyone sees any obvious mistakes I made in trying to make these maps Feat-compatible.
|