Hi, I am using 3dMEMA for doing ANOVA with repeated measures. Say I have two groups of subjects, GROUPA and GROUPB, and I have data from them PRE- and POST- intervention, so I set up 3dMEMA like below:

3dMEMA -prefix ANOVA_results -jobs 4 -groups groupA groupB -set groupA subA1 A1pre+tlrc A1post+tlrc subA2 A2pre+tlrc A2post+tlrc ... -set groupB subB1 B1pre+tlrc B1post+tlrc subB2 B2pre+tlrc B2post+tlrc ...

then the output will have 8 briks, including groupA:b, groupA:t, groupB:b, groupB:t, groupB-groupA:b, groupB-groupA:t, tau^2, QE:Chisq.

I want to look at the group differences in terms of their PRE-POST changes, so I look at groupB-groupA:t. However, I discovered that, if I switch the order of the input of time points, the results are very different, which is quite surprising. Can anyone help me to understand this? In the example I provided, does the output "groupB-groupA:t" represents group difference in terms of (groupB-groupA) on the time differences in terms of pre-post?

Thanks a lot!!


-Diane