Dear AFNI experts,

I have 2 groups of subjects: each subject performed a test with 4 conditions. First I created 4 regressors (using waver), and then I used 3dDeconvolve, giving for each subject a brick with least squares estimates of b0 (constant term), of b1 (linear trend), and of h1, h2, h3, h4 (my four regressors). All results look fine.
Next, I want to use 3dANOVA2 to compare activation between the two groups and I wonder what images to use for input. I would prefer to use the % signal change of (h2-h1, h3-h1 and h4-h1), corrected for each subject's own baseline. I have tried (waver amplitude)*(h2-h1) / b0, but the numbers I get are much smaller than I would expect (that is, << 1%). Any ideas what I am doing wrong?

A second question I have is whether it would also be sensible to use (for example) only the h2/b0 images as input image for 3dANOVA2, instead of pairwise differences between hx/b0 and hy/b0?

Thanks a lot,
Serge.