pingouin
pingouin copied to clipboard
Epsilon and Mauchly interaction in rm_anova2 differ from JASP
-
pg.epsilon
with two-way design gives slightly different but consistent GG and HF values -
pg.sphericity
do not work with two-way design: values are completely inconsistent.
Fixed for two-way interaction where one factor has only two values in e734df524be74504b6c66bb8dab41001bdb96573
In two-way repeated measures where each within factors have more than two levels, I think the regression approach is the only way. Need to create the design matrix using Patsy.