Error when running Friedman Conover test. about scikit-posthocs HOT 4 CLOSED

@Leicas I have updated the code. Please check if it is working now.

from scikit-posthocs.

No more error, some differences in the p-values but that was expected as it's a different calculation.

Thanks !

from scikit-posthocs.

@Leicas Thank you for your bug report. Please post a reproducible example (code + dataset).

from scikit-posthocs.

I unfortunately cannot share my data now but will try to make a reproducible example.

After investigating, the error is cased by a negative value of A, thus the sqrt fail.
Further inspection suggest that the (1 - T1 / (k * (n-1)) is the reason, T1=285.79 in my case, k is 4 and n is 12.

It seems that the implementation in https://github.com/cran/PMCMRplus/blob/master/R/frdAllPairsConoverTest.R is totally different than the old one in PMCMR, I'm trying to see if that would solve the issue but my knowledge of R is to limited...

from scikit-posthocs.