Discussion of "Robust Distance Covariance" by S. Leyder, J. Raymaekers, and P.J. Rousseeuw

Abstract: Distance covariance and distance correlation have long been regarded as natural measures of dependence between two random vectors, and have been used in a variety of situations for testing independence. Despite their popularity, the robustness of their empirical versions remain highly undiscovered. The paper named "Robust Distance Covariance" by S. Leyder, J. Raymaekers, and P.J. Rousseeuw (below referred to as [LRR]), which this article is discussing about, has provided a welcome addition to the literature. Among some intriguing results in [LRR], we find ourselves particularly interested in the so-called "robustness by transformation" that was highlighted when they used a clever trick named "the biloop transformation" to obtain a bounded and redescending influence function. Building on the measure-transportation-based notions of directional ranks and signs, we show how the "robustness via transformation" principle emphasized by [LRR] extends beyond the case of bivariate independence that [LRR] has investigated and also applies in higher-dimension Euclidean spaces and on compact manifolds. The case of directional variables (taking values on (hyper)spheres) is given special attention.