2024

Mrkvička, Tomáš; Myllymäki, Mari

Several new methods have been recently developed for the problem of domain selection when the test statistic is functional. That is, the aim is to distinguish between the two domains where the tested hypothesis should be rejected and where it should not. This work aims to compare these new methods through an extensive simulation study that consists of 1584 scenarios. The error rate, consisting of type I and type II errors, was chosen as a measure to compare the methods, regarding the two types of errors equally important. The false discovery rate methods consistently showed the best performance of all methods. When higher importance is given to type I error rate, the family-wise error rate methods should be preferred. From these methods, the global envelope test performed the best for smoother functions. None of the tested methods was consistently the best for very wiggly functions.