Probabilistic tests for the distributions in Rust's standard library. These are mostly significance level tests, and so will fail occasionally (assuming the distributions are implemented correctly... if they aren't then it will hopefully fail always).
Run with
rustc --opt-level=3 --test lib.rs && ./lib
This can take a long time, since it uses a lot of random
numbers. Adjust the constants in std_dists for more/fewer numbers,
higher is better (but slower).
- full of approximations (e.g. it replaces student-t by normal always)
- larger variety of tests (currently just performs t-tests on a set of sample means and sample observations to check these are close to the expected values)
- should probably use TestU01 and/or Diehard[er] instead of/as well as hand-written pure Rust tests.