On Testing Indipendence with Mixed-Type Data

Abstract

Here, we present novel independence tests for mixed-type data. In particular, we address the problem of testing independence between two multivariate vectors whose components consist of both continuous and count variables, as well as the problem of testing total independence within such vectors. We establish the limiting properties of the proposed tests and report the findings of extensive empirical studies. Furthermore, we discuss extensions of the results to high-dimensional settings.