Abstract

In this talk, we present new goodness-of-fit tests for parametric families of distributions on the hypersphere, based on Stein characterizations. We examine the asymptotic properties of these tests, including the limiting null distribution and their consistency. A resampling method is applied in order to provide an accurate p-value approximation. A comprehensive Monte Carlo simulation study demonstrates that our approach is competitive with the limited number of existing procedures in the literature. The results demonstrate that our tests not only provide competitive power across a range of alternatives but also offer robust performance across different distributional settings, outperforming existing methods in several cases.