0

I'm using the Kolmogorov-Smirnov test to test a random number generator. I'm wondering if there are other tests that I should implement as well. So far, I've looked at the equidistribution and serial correlation tests, but they seem redundant given that the KS already ensures that the random numbers follow the uniform distribution reasonable well.

Nick Stauner
  • 11,558
  • 5
  • 47
  • 105
Twilight Sparkle
  • 217
  • 1
  • 7
  • 1
    Hi, saw that earlier. IMHO, the question is slightly different because that question asks for the methods, not which ones are complementary with a KS approach. – Twilight Sparkle Apr 02 '14 at 08:25

1 Answers1

3

Serial correlation and the Kolmogorov-Smirnov test test different things, so they are complementary.

Consider the following silly example: If I take the rank of an observation and test whether that rank follows a uniform distribution, than that hypothesis will not be rejected. Would that mean that taking the rank is a random number generator? Of course not, and the problem will be obvious when you look at serial correlation.

Maarten Buis
  • 19,189
  • 29
  • 59