Power Analysis for 3 Groups and 3 Measures

My study “needs to be powered to detect individual group comparisons, not simply the omnibus ANOVA because this is where the hypotheses will ultimately be tested” (feedback I received from a proposal reviewer). The hypotheses is that group membership will be predicted by score on outcome measures being used.

I had initially ran my power analysis based on an omnibus anova for three groups using three measures, however, it seems as though this is not correct. Does anyone have any ideas as to how to better meet the reviewers comment?

When you have 3 groups that are of roughly equal priority, the reviewer is quite incorrect, and the overall power from ANOVA is a primary quantity to use. But better is to not use power at all and compute the expected precision of any contrast of interest (e.g. margin of error; half-width of 0.95 compatibility interval).

1 Like