Gang Chen wrote:

> Yes, that is right. However, you can use another more amenable
> option -gltsym to simplify the specification.
>
> > It seems this matrix is specifying separate comparisons at
> each
> > time point of the lags, when all I need is a single auc
> comparison for
> > the LC, and a dumpout of the HRF response to each -stim_file
> class.
>
> I assume the big matrix is for comparing two conditions. There
> are two approaches to set up such a contrast: (1) Comparing the
> condition at each time point; (2) Comparing their UAC. The two
> mean differently and you would come up with two different
> results, and it is your choice.
>
> Gang

Hi Gang, I am content to simply make an omnibus comparison of AUC in my linear contrasts, where the one-line matrix file is used. I gather then, that

1) using -iresp to dump out % signalchange requires a stipulated duration of maxlag back in the -stim_file specs, and

2) specifing the maxlag requires expanding the one-line matrix somehow to accomodate the maxlag spec.

However, the script I posted and the expanded one-line matrix I suggested results in an analysis with a collinearity warning, and a bucket dataset that is capped at 0-255 subbricks, because the -coef and t- subbriks of each of the 124 items in the expanded matrix uses up 248 of them! This is clearly not the solution I am looking for.

should my matrix instead go from:

0 0 0 0 0 0 -1 0 1 0 0 0 0 ... etc

to

0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 .... etc?

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.

Jim