search

brentp / fishers_exact_test

Fishers Exact Test for Python (Cython)
BSD 3-Clause "New" or "Revised" License
62 stars 21 forks source link

Tests failing on Python 3.3 #7

Closed johanneskoester closed 8 years ago

johanneskoester commented 9 years ago

Hi, as the title says, your rfisher.py tests are currently failing with Python 3.3. The calculated p-values do not even remotely match the expected ones. Is there anything I can do to help with debugging?

brentp commented 9 years ago

I think this is related to 784b16c13f93dc7f9501e75709bfcec28e981e81 . You could try setting epsilon = 1e-10 For me, that makes the tests pass. I don't remember why this is needed and don't want to keep changing back and forth until I understand what's going on (though 1e-10 does seem to give more consistent results).

johanneskoester commented 9 years ago

Well, perhaps you run into numerical issues with the smaller epsilon?

johanneskoester commented 9 years ago

Am I right that the change was introduces because of issue #1? If so, maybe there is a compromise between the two extremes such that yours and the test case of issue #1 succeed?

brentp commented 8 years ago

@johanneskoester the latest version has passed these tests. Let me know if you see any problems.

johanneskoester commented 8 years ago

Great, thank you!