Goodness of fit tests in SciPy

Question

I'm new to Python and coming from the R world. I'm trying to fit distributions to sample data using SciPy and having good success. I can make distribution.fit(data) return sane results. What I've been unable to do is create the goodness of fit statistics which I'm used to with the fitdistrplus package in R. Is there a common method for comparing "best fit" from a number of different distributions with SciPy?

I'm looking for something like the Kolmogorov-Smirnov test or Cramer-von Mises or Anderson-darling tests

For which distributions to you need the gof tests? With estimated parameters the distribution of the test statistic is different from the one with fully specified distributions. statsmodels, see John D Cooks' answer, has more in the sandbox and should be available by the end of summer. — Josef, Jul 02 '12 at 09:58

score 17 · Answer 1 · answered Jun 29 '12 at 20:41

17

See the scipy.stats library: http://docs.scipy.org/doc/scipy/reference/stats.html

It contains K-S and Anderson-Darling, although apparently not Cramer-von Mises.

answered Jun 29 '12 at 20:41

BrenBarn

242,874
37
412
384

1

many thanks! I wish I could accept both your answer and John's. – JD Long Jul 03 '12 at 16:11
Cramer-von Mises test is available with scipy.stats: https://docs.scipy.org/doc/scipy/reference/stats.html#statistical-tests – varsh Apr 27 '22 at 07:02

score 15 · Accepted Answer · answered Jun 29 '12 at 21:57

15

There's also statmodels goodness of fit tests.

answered Jun 29 '12 at 21:57

John D. Cook

29,517
10
67
94

Goodness of fit tests in SciPy

2 Answers2

Linked