posthoc contrasts from an interaction

[Fabio]

dear aaron, dear glmflex gurus,

I am trying to compute a posthoc contrast for a mixed model specified like this:
I.Model = 'fb*group + random(participant_id|fb)';

where fb is a within subject factor. I want to test whether the difference in my dependent variable (call it Y) between the two levels of 'fb' (call them 1 and 2) differs between groups (call them A and B).
Namely I want to test whether:

[y(A,1)-y(A,2)] - [y(B,1)-y(B,2)] > 0


I tried (delirating attempt I know ;-) ):
group$A|fb$1 & group$B|fb$2 # group$A|fb$2 # group$B|fb$1

but results do not make sense. Also reading other topics in this group it looks like this contrast cannot be computed within GLMFlex, but I would like to make sure.

Thanks
again

Fabio

[Aaron Schultz]

Post-hocs can only be used to make simple comparisons, that is comparisons between two conditions or one way ANOVAs across multiple conditions.  If you are trying to test a "sub interaction", there is not currently a way to specify that contrast with GLM_Flex_Fast.  I would recommend re-running the model with only the data for the sub-interaction and then look at that interaction map.  The error term will be slightly different, but should be pretty close.

-Aaron