Dear Christian,
Thanks for your quick response, I really appreciate.
(1) I realised that my previous message was not clear enough. My apologies.
Reminder, my flexible factorial is: Subject, Group (between; 2 timepoints) and Time (within; 2 timepoints).
As far as I understood from the paper, the inflation of the F statistic applies to the Main effect for group (between-subject effect), but not to the group*time interaction or the main effect for time.
Primary aim (A) is the Group*time interaction and if null cannot be rejected then test for the main effect for time (within)(secondary aim; B).
(A) Group*time interaction. If usual statistics rationale applies, to test for this effect I would need to adjust for covariates like TIV, age and gender to correct for imbalances at baseline and to increase power by modelling the variability in outcome (in this case, GMV) explained by relationships with these covariates. However, when I run my flexible factorial with nuisance I see that the covariates have not been taken into account (exact same results as with no nuisance covariates specified in the model). But I see from a previous post, that it seems that SPM cannot handle it:
https://www.jiscmail.ac.uk/cgi-bin/wa-jisc.exe?A2=ind1606&L=SPM&P=R64637&X=B62D057C9FBB2A6716&Y
Should I use the SwE or it is not worthy for this design with only 2 timepoint (pre, post)?
(B) Main effect for Time: no covariate adjustment is needed because we focus on intra-individual effects.
(2) That is clear thanks!
Again, thanks so much in advance to Christian and the CAT12/SPM community,
Kind regards
Carlos
|