Local polynomial regression

This thesis examines local polynomial regression. Local polynomial regression is one of non-parametric approach of data fitting. This particular method is based on repetition of fitting data using weighted least squares estimate of the parameters of the polynomial model. The aim of this thesis is therefore revision of some properties of the weighted least squares estimate used in linear regression model and introduction of the non-robust method of local polynomial regression. Some statistical properties of the local polynomial regression estimate are derived. Conditional bias and conditional variance of the local polynomial regression estimate are then approximated using Monte Carlo method and compared with theoretical results. Powered by TCPDF (www.tcpdf.org