Dear Daniel,
> I've been trying to estimate a model with 10 subjects, each of
> which has 8 runs of 176 TRs each. However, I ran out of memory when model
> estimation was 29% complete. So, it looks like I can't model estimate
> all 10 subjects together. I know that I can model estimate a smaller
> groups of subjects. Then, I could generate a contrast image for each
> subject. And, then, I could do basic stats on a series of contrast images
> to do a random effects model. I have 2 questions, though.
>
> 1) If one of my contrasts is an interaction (and I have an interaction
> contrast image for each of my 10 subjects), how do I analyze that at the
> second level?
This is very simple - you use a one sample t test under basic models.
This tests the null hypothesis that your interaction effect is 0.
> 2) Is there any way to do a fixed effects model with all 10
> subjects, even though I can't estimate the model with all 10 subjects
> together?
Reduce the data by subsampling to larger voxels and ensure you are
using the latest version of SPM99?
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|