I ran a structural analysis in randomise and obtained a main effect (ME) of group and also an group x measure interaction in two regions, both tfce-corrected p<0.05. The location of the ME and the interaction overlap although the cluster sizes vary, with the cluster size of the ME being larger in both instances (Cluster 1 = 174 voxels in group ME, 73 voxels in group x measure interaction; Cluster 2 = 185 voxels in group ME, 173 voxels in group x measure interaction). I understand from Jeannette's page that a ME of group is difficult to interpret in the presence of a sig. group x measure interaction.
What's the best way to interpret the group ME data, especially in the case of Cluster 1, where a sizeable portion of the cluster is not overlapping with the interaction? Is it appropriate to interpret the 101 unique voxels in the Cluster 1 ME as representing a ME of group but no group x measure interaction? Could the same be said for Cluster 2 or are the 12 unique voxels likely to reflect noise?
Thank you in advance for your help,
Cindy