Twenty minutes before the talk coffee and cookies will be available in
Bear Hall 211.
Title: Function Estimation via Kernel-based Methods.
Speaker: James Blum
Abstract:
An introduction to estimation of functions using non-parametric methods,
focusing on kernel-based methods for estimating bivariate relations and density
functions for a single, continuous random variable. Asymptotic properties
of these estimators will be explored and related to basic issues in kernel
function and smoothing parameter (bandwidth) selection. Further, methods
for overcoming local bias of the estimator will be reviewed.