randhy

Hypothesis thesting with approximate randomisation

Approximate randomisation is a significance testing approach suitable for NLP problems.

🤔 Why not a traditional t-test?

While randomisation tests are just as good as analytical approaches such as the t-test, they are better when the assumptions of the latter are not met and they are also quite simple to implement.

🖥️ Installation

pip install randhy

References

1. William Morgan, Statistical Hypothesis Tests for NLP - Stanford Computer Science (slides)
2. Wassily Hoeffding. 1952. The Large-Sample Power of Tests Based on Permutations of Observations. Annals ofMathematical Statistics, 23, 169–192.
3. Eric W. Noreen. 1989. Computer Intensive Methods forTesting Hypothesis. John Wiley & Sons