I have a mixed anova with repeated measures design: two between-subjects factors, doctor (3 levels) and method (3 levels), and one within-subject factor, time (1w, 3w, 2m, and 6m).
3 doctors would be randomly selected, so "doctor" would be random effect in addtion to "subject". Both methods and time are fixed effects.
I read something about effect size for mixed anova. It seems that many people would like to use eta-squared. For this study, the effect size is more or less like Cohen's d. In particular, I would like the mean of Method A to be at least 3 units greater than Method B, 4 units greater than C. So how should I calculate the effect size?