Hi Liz,
Actually the part I feel confused with is the way you did the percentage change. My impression is that people usually calculate the percentage of the beta coefficients after they run 3dDeconvolve. They usually add the option "-peak 1" when they run waver to generate their regressors as "stim_file" inputs in 3dDeconvolve. When they calculate the percentage change of the beta coefficients, there is no need to scale since the basis functions of gamma variate is one.
If I understand you correctly, you did the percentage change of the original signal before (instead of after) you run 3dDeconvolve. It seems to me that after you run 3dDeconvolve, you would still come up with the same beta coefficients (and baseline parameters as well) whether you did the % signal change or not. And it is usually the % change of the beta coefficients that people run group analysis on.
Gang