Executive Summary : | This proposal consists of four main components: kernel-based non-parametric regression, shape-restricted non-parametric regression, quantile-based methodologies, and measure of association/test for independence. It aims to compare misaligned regression curves, analyze variable selection in multivariate non-parametric regression models, and study measurement error in non-parametric regression. The proposal also plans to study the asymptotic properties of least squares estimator when multivariate nonparametric regression functions are misspecified, and work on monotone regression when covariates are infinite-dimensional. The proposal also explores quantile issues in signal processing models, specifically the local polynomial quantile estimator of unknown regression functions. The proposal also aims to test for independence for more than two random variables. |