Best representation of group stats

AFNI Message Board
1

Hello,

I have data from a 2x2 repeated measures design (Time: Pre, Post, Condition: A, B) and I performed a group level analysis using 3dANOVA3 -type 4 -alevels 2 -blevels 2.

My question is how do I best display which regions showed a significant difference with respects to group and time and in what direction? Is it appropriate to create a mask based on the time_cond_interact map and then use 3dcalc to look at PrevsPost difference for the A and B condition separately only within the significant interaction mask? Or should I use the condition effect map as a mask and then use 3dcalc to look at PrevsPost difference only where there was a main effect of condition? Or is there another option that is more appropriate?

Thanks

2

In the conventional statistics textbook, there is something tangentially related to what you’re describing: for each main effect or interaction, only look at those post hoc tests after checking the omnibus F-statistics. The rationale is to have some extent of weak control for the multiple comparisons issue. Don’t bother, and forget about the masking ritual. In your case just go directly to each effect or contrast and check it out.

3

Hi,
Try this for ideas.
It will plot interactions to see direction etc.
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/tutorials/ClusterExplorer/Introduction.html

thanks, Justin