> Can you please tell me how 3dDeconvolve finds the beta weights
> without caring about the amplitude difference between the
> regressors and the data?
The beta vector (of all regressors) in a regression model Y = Xb + e is basically solved as
b = (X'X)^-1 X'Y
The columns in design matrix X are the regressors. So if you scale one regressor by a factor c, the above formula indicates that the beta for this scald regressor would be c times the original beta value.
Gang