I have been researching ways to calculate the necessary sample size to detect moderators - gender (2 levels) and education (3 levels). The study has two primary groups with 6 repeated measures. The primary outcome of interest is test score. We want to include gender and education in the model and determine if they are moderators. I have been looking at PASS and Gpower software, without much luck and from what I've been reading, a simulation analysis may be my best bet. I have a general idea of what the distribution of gender and education would be within the two primary groups, but do not know what the possible test scores may be across the two groups by gender and education, since there is no previously published work. I was wondering if anyone could provide guidance or references they've found useful that may help me determine what else I need in order to calculate the necessary sample size.