CONF keller:nips:2005/IDIAP Benchmarking Non-Parametric Statistical Tests Keller, Mikaela Bengio, Samy Wong, Siew Yeung EXTERNAL http://publications.idiap.ch/attachments/papers/2005/keller-nips-2005.pdf PUBLIC http://publications.idiap.ch/index.php/publications/showcite/keller:rr05-38 Related documents Advances in Neural Information Processing Systems, NIPS 18. MIT Press 2005 IDIAP-RR 05-38 Although non-parametric tests have already been proposed for that purpose, statistical significance tests for non-standard measures (different from the classification error) are less often used in the literature. This paper is an attempt at empirically verifying how these tests compare with more classical tests, on various conditions. More precisely, using a very large dataset to estimate the whole ``population'', we analyzed the behavior of several statistical test, varying the class unbalance, the compared models, the performance measure, and the sample size. The main result is that providing big enough evaluation sets non-parametric tests are relatively reliable in all conditions. REPORT keller:rr05-38/IDIAP Benchmarking Non-Parametric Statistical Tests Keller, Mikaela Bengio, Samy Wong, Siew Yeung EXTERNAL http://publications.idiap.ch/attachments/reports/2005/keller-idiap-rr-05-38.pdf PUBLIC Idiap-RR-38-2005 2005 IDIAP To appear in Advances in Neural Information Processing Systems, NIPS 18. MIT Press, 2005. Although non-parametric tests have already been proposed for that purpose, statistical significance tests for non-standard measures (different from the classification error) are less often used in the literature. This paper is an attempt at empirically verifying how these tests compare with more classical tests, on various conditions. More precisely, using a very large dataset to estimate the whole ``population'', we analyzed the behavior of several statistical test, varying the class unbalance, the compared models, the performance measure, and the sample size. The main result is that providing big enough evaluation sets non-parametric tests are relatively reliable in all conditions.