Hi all,
I've got a question about how to choose the various parameters needed by tent functions when using 3dDeconvolve to analyze data with irregular stimulus timing (as outlined in educational materials afni07.pdf, see [
afni.nimh.nih.gov]).
We have a FMRI paradigm in which people are learning associations between visual cues and responses. We have reason to believe that the time course of activation will change with learning. So we would like to be able to model changes in the time course of brain activation with learning, rather than use fixed-shape gamma functions as ideal references functions. Unfortunately, we designed the experiment expecting to use gamma functions. The stimulus timing is completely irregular, not locked to the TR, so we are trying the approach outlined in afni07.pdf.
The tent functions in the example PDF are tent(x/4), tent((x-4)/4), tent((x-8)/4). What factors go into the choice of the the denominator parameter, in this case 4 and the subtraction parameter, in this case 0, 4, and 8?
My scanning parameters were TR 3.6 sec, and I expected from a previous experiment that the hemodynamic response function in this experiment would peak at 5.5 sec. I used the following logic to create three tent functions. I chose one tent function to be centered on the expected peak time, one centered 1 TR previous and a third centered 1 TR later as follows: tent((x-1.9)/3.6), tent((x-5.5)/3.6), and tent((x-9.1)/3.6). Are these selections any more or less valid than those chosen in the example PDF? What criteria ought to go into this decision?
Thanks for any info.
-jim