> Do I convolve these two regressors with an HRF first? Or can I use minlag and maxlag and
> then calculate an area under the curve?
I cannot see how this could be done with the deconvolution approach using minlag and maxlag: maybe 3dDeconvolve takes -1s with minlag and maxlag? So either assume an HRF, or use -stim_times_AM2 with stimulus onset timings with TENT if you still want to run deconvolution (and use -gtlsym to get the area under the curve for the 2nd regressor).
> for the second regressor, it would really only be coding "-1' for condition 1 and "+1" for
> condition 3. Coding "0" for condition 2 would be indistinguishable from other trials that are
> coded with zeros, like baseline trials (or any trials that are neither conditions 1, 2, or 3).
> Isn't this the old problem of effect coding with three conditions, where the middle condition
> is essentially ignored?
No, it would be fine in this case because, similar to what you've done in 3dRegAna, there is an "intercept" -- the first regressor -- that basically treats condition 2 as a reference level.
Gang
Edited 1 time(s). Last edit at 04/14/2014 05:44PM by Gang.