I am trying to do some post hoc analyses on a group-level melodic job that contains data from two groups of subjects. The fMRI data are task-based (block design, one single run per subject). In the melodic job I have included the first-level EVs (design.mat and design.con) representing the task, as well as group-level EVs and contrasts, to show group differences. When doing this and looking at the Sessions/Subjects Mode output to see group differences for a single IC, I see plotted the per-subject data representing “subject/session mode” and “full model fit”. My questions are:
1) What exactly is the “full model fit” in this context, and how is it calculated?
2) In the statistical analyses (GLM OLS) that follow, what exactly is the rationale for using the "model fit" as the dependent variable, as opposed to the “subject/session modes”?
Thanks - John
|