Invalid permutation test ?

[trevismd]

Hello,

I was glad to find an implementation of permutation tests to challenge my theoretical understanding of it (and compare with my own).

Actually, it seems that mlxtend.evaluate.permutation_test wrongly rejects permutations whose statistic result is equal to the tested statistic (exact tests here and simulations here), which leads to lower p-values than desired.

The fix is quite simple. Should I open a PR ? Or did I miss something ?

Thanks

[rasbt]

Good catch and thanks for the note. It should indeed be "greater or equal to" (>=). If you'd like to open a PR, that'd be nice.