Hello SPM users,
I was wondering if I could tap in to your collective wisdom:
A colleague wishes to use PPI in SPM8 to explore a straightforward contrast [A vs NOT_A] on some data acquired over 10 runs. She has been following along with the SPM8 manual (Chapter 33), where it suggests concatenating the runs into a single super-session. When she does so, and performs a GLM on the concatenated super-session, she notes that the A vs NOT_A effects are noticeably diminished (compared to when 10 individual runs are entered into the GLM analysis), and is naturally concerned that this will adversely impact the sensitivity of the PPI analysis. I assume the apparent reduction in the magnitude of the observed effects are related to mathematical quirks (e.g., things like degrees of freedom) and an increased intra-session variability (all the between-session fluctuations now appear within the super-session).
Given that the PPI analysis can include block regressors, is this a cause for concern? If so, what is the appropriate course of action?
Thanks