Comparing gPPI Results Between Two Groups: Addressing Baseline Connectivity Differences

AFNI Message Board
1

Dear AFNI experts,
I am hoping to get your help about comparing gPPI results between two groups.

I had two groups of participants: one healthy group and one patient group, and each group had two conditions, A and B. I calculated the gPPI results for each group in each condition.

I wanted to know how to directly compare the differences in gPPI results between the two groups within the same condition.
When I simply subtracted the gPPI results for the same condition between the two groups, the gPPI results for the healthy group were always smaller than those for the patient group.
However, when I separately subtracted the gPPI results for conditions A and B for each group, and then subtracted the difference between the healthy group's A and B results from the difference between the patient group's A and B results, the gPPI results for the healthy group were greater than those for the patient group.

I suspected that significant baseline connectivity differences existed between the two groups, which made it inappropriate to simply subtract the gPPI results for the same condition when comparing the two groups. I was looking for a better method to compare the differences in gPPI results between the two groups within the same condition.

Thank you so much in advance.
Ting

2

Ting,

When I simply subtracted the gPPI results for the same condition between the two groups, the gPPI results for the healthy group were always smaller than those for the patient group.

This first analysis reveals group differences within each of the two conditions individually

However, when I separately subtracted the gPPI results for conditions A and B for each group, and then subtracted the difference between the healthy group's A and B results from the difference between the patient group's A and B results, the gPPI results for the healthy group were greater than those for the patient group.

This second analysis examines group differences in the contrast between the two conditions.

These two analyses may not necessarily exhibit the same patterns, particularly in the presence of an interaction between group and condition, as indicated by the second analysis. To further disentangle these effects, you can perform a 2 \times 2 ANOVA and parse various effects using -gltCode in 3dMVM or 3dLMEr.

Gang Chen

3

Thank you so much, I will try. :blush: