A0860
Title: Large sample behavior of high-dimensional Kendall's correlation matrices with dependence
Authors: Priyanka Sen - Indian Institute of Technology Bombay (India) [presenting]
Monika Bhattacharjee - IIT Bombay (India)
Abstract: The limiting spectral distribution (LSD) of Kendall's correlation matrix has been established in the literature under conditions of row dependence. However, the assumptions used in prior works are non-interpretable and difficult to validate, especially in non-Gaussian settings. The same is established under more interpretable assumptions that are easier to visualize and can be validated in a wider range of non-Gaussian models. Moreover, the joint convergence of multiple Kendall's correlation matrices is established, and the LSD of polynomials is formed from these matrices. This significantly extends the existing results by providing a more general framework for studying the spectral properties of correlated data.
