Assuming you want to take one sub-brick of each subject (such as a coefficient in the stats file output by 3dDeconvolve), then
3dTcorr1D can do this. Though you'll have to do some basic legwork to get the files ready. First, put all of your scalar values in one text file (.1D), each column representing a different variable. Then you can use 3dbucket to create a single file with all of the 15 subjects (my example shows pulling out "condition1" from each file):
3dbucket -prefix AllSubjects Subject1+tlrc'[condition1#0_Coef]' Subject2+tlrc'[condition1#0_Coef]' Subject3+tlrc'[condition1#0_Coef]'
Finally you can run 3dTcorr1D:
3dTcorr1D -prefix MyCorrelationMap AllSubjects+tlrc ScalarData.1D
You will be able to threshold the resulting image in AFNI to see which voxels are significant at different r-values. You'll also notice that the program correlated all 15 variables with your image data and can scroll through them.
It's worth noting that you can also do this in 3dttest++, 3dMEMA, 3dMVM, and 3dRegAna. But I think the 3dTcorr1D is likely the most straightforward of these approaches for data like you describe.
Edited 4 time(s). Last edit at 09/23/2014 06:58PM by Peter Molfese.