Hi all,
I have what I am sure is a simple-to-answer question.

I have a model of my fMRI task in which certain events of interest for a PPI analysis only have onsets falling in earlier time-points of the experiment (but the later time-points do comprise events of interest for other analyses, hence their inclusion in the model). If I follow the standard PPI procedures - pull a VOI timeseries from a region in the brain using my "standard GLM" SPM.mat file, and then generate a PPI file by setting up a contrast between my events of interest - I wind up with a timecourse that extends beyond the time-points of interest for my PPI. As a consequence, when I specify a PPI design for model estimation I wind up with a PPI.ppi regressor with values loaded on the first half of the timeseries, and PPI.Y with values encompassing all BOLDs, and a PPI.P with values again restricted to the first half. 

On the surface, the method seems designed to work with a situation where only some of your time-points fall under the umbrella of the chosen events of interest, but I'm a bit concerned about having a design with this particular structure. Will this throw the results off in some way? If so, I take it my only option is to specify a new "standard GLM" model to get an SPM.mat where the second half of the time-points are not included, and thus my PPI with early-phase events I won't wind up with this mismatch in length for the .ppi, .Y, and .P regressors


Thanks for your guidance!
Thackery