Dear SPMers

I conducted an event-related fMRI study where I have 2 groups of adults performing 3 event-types . The two groups differ in age, and therefore there are several caveats for interpreting between group main effects: differences in baseline activity and signal variance to name a few. As such, I decided on focusing on ROIs defined from within group RFX analyses (from both young and older adults RFX) and then conduct post-hoc  ANOVAs on these functionally defined ROIS to identify activations that exhibited group-by-task interactions and then plotted the results for significant ROIs. I felt at the time this was the best way to circumvent the issues concerning age-related differences in BOLD response and possible between groups differences in variance and baseline activity.  However, I received a review requesting I directly compare both age groups. Traditionally I have used multivariate methods to look at between group effects & am just starting to use SPM2 for this type of analysis so  I hope my questions/concerns are not too naive.

To answer the reviewer’s request I plan to do the following & just wanted to get confirmation if the following methods are ok, even though signal:noise  may differ between groups.  To answer the question of what activations are 1) common between groups  2) exhibit groupXtask interaction 3) show a group main effect (which the reviewer requests,  but I have problems with for theoretical reasons) I plan to  enter con*img from both age groups into separate one-way ANOVAs for contrasts/effects of interest.

My concerns are:

- how is the variance calculated & is the correct error term applied (i.e.concern about unequal variance)

-if there are baseline activity differences between group - will this mean I won't see any effects? or limited effects?

-also will this allow me to also examine groupXtask interactions?

-are the results generalizable (RFX) ?

Regards,

Natasha