Hello,
I've been reading through many previous posts on thresholding, and I still have some confusions about how to apply thresholding to my data. Sorry for the perhaps repetitive questions, just trying to wrap my head around this stuff.
I've run a linear regression using 3dDeconvolve. In that command I created a number of general linear tests (GLTs), e.g.:
-glt_label 1 person_delay -gltsym 'SYM: +0.5*P.ED +0.5*P.LD'
..where I'm interested in the average brain response to the P.ED and P.LD regressors within a SINGLE SUBJECT.
Now, the output bucket from the regression contains a Coef, T-stat, and F-stat sub-brick for this GLT. My intention now is to discover which voxels are significantly "activated" by the conditions described by the GLT. I can overlay the T-stat sub-brick for this GLT and adjust the threshold bar to change the threshold, but what is the "critical" threshold to set this at so that I have some confidence these voxels are above-threshold voxels for this task?
Isn't there a way to just determine what the "critical threshold" of my contrast is and then adjust the threshold slider when looking at the GLT.T-stat sub-brick? Or is this the point where I need to do some sort of cluster analysis on the data (e.g. using 3dFWHMx , AlphaSim, and 3dmerge, etc.) to find significant "clusters" of voxels and then perform further statistics on those clusters and create new output sub-bricks with which to use as overlays? I'm confused about where to go at this step.
I want to be able to say "these voxels in the brain of SUBJECT X were significantly activated by task Y", which is a rather basic question, but I'm getting stuck on using the correct procedures to address that question.
Thanks for any direct feedback (or pointers to a description of this analysis).
Jarrod