> We then report only the clusters that are significant from the t-tests and
> corresponds to the significant clusters we saw in the whole brain interaction F
> map. All the significant clusters that we report in tables have the same cluster
> thresholding value based on alphasim.
>
> Is this a good approach? or am I missing something?

Did the family-wise error correction with AlphaSim apply to the F- or t-map, or both?

> Reviewers of our paper (who seem to use SPM) are saying that t-tests should
> be not whole brain but should be done only on clusters that were significant in
> the interaction map and different p values should be given for every cluster.

I'm confused here: Do you mean the reviewers suggest that only statistical analysis at ROI level (averaged across the ROI?), but not the whole brain, is legitimate?

Gang