Change-point analysis for matrix data: the empirical Hankel transform approach

Abstract

In this study, we introduce the first-of-its-kind class of tests for detecting change-points in the distribution of a sequence of independent matrix-valued random variables. The tests are constructed using the weighted square integral difference of the empirical orthogonally invariant Hankel transforms. The test statistics have a convenient closed-form expression, making them easy to implement in practice. We present their limiting properties and demonstrate their quality through an extensive simulation study. We utilize these tests for change-point detection in cryptocurrency markets to showcase their practical use. The detection of change-points in this context can have various applications in constructing and analyzing novel trading systems.