Hi Ian,

The original COPEs are brought up because the variance between subjects is then preserved. Using z- or t-stats would destroy this information and would not be appropriate. Intra-subject variances are a different story: they are much smaller than between-subject variances and, notwithstanding, we hope to have something in the future months to approach this in a proper way. Regardless, it's fine and well appropriate use simply the COPEs in randomise.

All the best,

Anderson

On 25 September 2015 at 05:02, Ian Ballard <[log in to unmask]> wrote:

Hello,

I had a quick question about inputs to randomise for fMRI. I know that normally you use the second-level COPEs as inputs. I was wondering whether it would be appropriate to instead use the t- or z-stat images from the second-level fixed effects analysis. It seems to me like this would allow you to account for differences in intra-subject variances when you do the permutation test. Please correct me if I'm wrong or missing something...

Thanks,

Ian