> the script I posted and the expanded one-line matrix I suggested
> results in an analysis with a collinearity warning
There are a couple of causes for collinearity. One possibility is related to the lags you choose, another is by mistake in setup. So you'd better pinpoint what is causing the problem. After that you can make some adjustment. My maximum lag of 7 sounds a little too big: What is your TR?
> All I want is to retain my old tried and true GLT-"AUC"
> 3dDeconvolve analysis using canonical HFR vectors, but add the
> functionality of -iresp to dump out a dataset for each stimulus series
> which gives me a voxel-based output that shows the percent signal
> change for each of the HRF timepoints at and following stimulus
> presentation. Are these incompatible objectives? It seems that they
> shouldn't be.
I get lost here. What you do with constrast setup should not have anything to do with the output of -iresp. What is your concern?
Gang