7

I've been doing some data analysis with Scipy. So far I accomplished this with continuous distributions:

  • I can fit a probability distribution to a set of data points using a maximum likelihood fit. For example using stats.chi2.fit(data_points).
  • I can test if my samples comes from a specific distribution. For example using stats.kstest(data_points, 'norm')

I tried to do the same analysis with discrete distributions (like Poisson, Binomial and Geometric), but it seems like Scipy doesn't offer similar functionality for discrete distributions. What's the standard approach for fitting and testing for discrete distributions?

Silverfish
  • 20,678
  • 23
  • 92
  • 180
Carlos Gavidia-Calderon
  • 143
  • 7
  • Welcome to our site! Have a look at our [help/off-topic] to see what kinds of subject are on- and off-topic here. The question as it's currently written seems to have a heavy focus on a software implementation ("how do I accomplish those tasks using Scipy" is generally off-topic). But it seems to me that you may well have an underlying statistical issue which is on-topic here. Do you think you could edit the question to bring this to the fore? – Silverfish Mar 02 '16 at 10:05
  • @Silverfish thanks for the suggestion. I did the edit to focus on the statistical issue – Carlos Gavidia-Calderon Mar 02 '16 at 10:23
  • Thanks, this looks better to me, though it may well be a duplicate. – Silverfish Mar 02 '16 at 13:00
  • Try this search by relevant tags: http://stats.stackexchange.com/questions/tagged/discrete-data+goodness-of-fit – Silverfish Mar 02 '16 at 13:02

0 Answers0