A Statistical Significance Test for Person Authentication
Samy Bengio, Johnny Mariéthoz · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 2004
Assessing whether two models are statistically significantly different from each other is a very important step in re-search, although it has unfortunately not received enough at-tention in the field of person authentication. Several perfor-mance measures are often used to compare models, such as half total error rates (HTERs) and equal error rates (EERs), but most being aggregates of two measures (such as the false acceptance rate and the false rejection rate), simple statisti-cal tests cannot be used as is. We show in this paper how to adapt one of these tests in order to compute a confidence in-terval around one HTER measure or to assess the statistical significantness of the difference between two HTER mea-sures. We also compare our technique with other solutions that are sometimes used in the literature and show why they yield often too optimistic results (resulting in false state-ments about statistical significantness). 1.