> I have a meaningful interaction involving two groups for a contrast of Song type A > Song type B.
Is the interaction between Group and Song type based on the whole-brain (or voxel-wise) analysis?
> I would like to now plot the betas for this interaction including the lower order components (song
> type A > baseline and song type B > baseline for each group separately) to better understand the
> interaction but I wanted to ask what would be the most appropriate way to handle the covariate
> related to Song type B?
Plot out at the ROI level?
> Regress it out of the betaweights for all of the song types >baseline conditions (collapsed across
> groups)? Only the Song type B > baseline conditions (since the covariate only applies to this
> condition) collapsed across groups?
Run ANCOVA (two-sample t-test plus a covariate) with 3dttest++ on Song type B.
> The covariate is mean centered across groups.
Centering is a delicate issue. Do the two groups differ significantly in the average value of the covariate? If centered properly, you may extra the intercept from the output of 3dtttest++ (which would correspond to the effect of Song type B *controlled* at some center value(s)), and then do the plotting. Alternatively you may do the ANCOVA step at the ROI level too if you prefer.
Gang
Edited 1 time(s). Last edit at 03/02/2015 11:16AM by Gang.