Power calculation for multicategorical predictor in longitudinal data

[Matt_Southward]

I'm curious if anyone has thoughts about how to set up a power simulation. I have a sample of 70 people who have completed 12 sessions of therapy, which consists of 5 different skill modules delivered in a random order. With proc mixed, I used an indicator variable representing the skill modules as a predictor of continuous depression symptoms at each session. The Type 3 test of fixed effects for the indicator variable was significant - specifically, depression was significantly higher during one module compared to the others. I'm not sure, though, how to set up a simulation with a multicategorical predictor since SAS seems to calculate the Type 3 tests automatically and separately from the modeling input. For reference, I'm most familiar with Lane and Hennes' power code, which covers continuous and binary predictors but not categorical ones: https://journals.sagepub.com/doi/suppl/10.1177/0265407517710342

[Ksharp]

I remembered @Rick_SAS  wrote a blog about this topic before .

[Rick_SAS]

The Lane and Hennes paper appears to be behind a paywayll, so cannot access.

Show your PROC MIXED code. It is not clear how you are handling the longitudinal aspect of the design. Also, sample data is always appreciated.