Hi
In my group level model, I used a behavior score as a covariate (demeaned and then entered in FEAT) and I wanted to know if the activation within a precuneous ROI mask was correlated with the behavior measure during a task. Small-volune correction (Randomise) showed a significant cluster within precuneous, and I obtained the peak coordinates by using Cluster. Then I wanted to illustrate the correlation. I thus used two methods to extract the response in the peak voxel: Featquery and Fslmeants. It was supposed to have a significant correlation if I regressed the extracted values against the raw behavior score (i.e., not demeaned score). Surprisingly, although the correlation with data extracted via Fslmeants was as expected highly significant (p=0.002), those obtained via Featquery exhibited only trend correlation (p=0.081). Why? (Suppose Fslmeants gives me parameter estimates, and Featquery returns percent signal change?)
Thanks. Mike
|