Sphericity (Mauchly) test inconsistency

[plankter]

After performing repeated measures ANOVA, I discovered that in some cases the output of pingouin.sphericity function differs from the output of R/other apps.

pg.sphericity( data=df, dv="Value", within="Bin", subject="Animal", method="mauchly", )

Here are some basic examples:

• Correct (similar results):

Correct.csv

• Incorrect:

Incorrect.csv

After uncommenting Method 1 implementation, results looks similar to R in all cases, see here:

pingouin/src/pingouin/distribution.py

Line 980 in cfaa5db

# Method 1. Contrast matrix. Similar to R & Matlab implementation.

Best regards,
Anton

[raphaelvallat]

Thank you for opening the issue. This should be fixed in #482