Local Regression and Liklihood.

Local regression is a method for smoothing data in one or more dimensions. The three classes will introduce the method and survey recent developments and extensions to likelihood models and applications. Model selection procedures such as cross validation will also be discussed. The presentation will emphasize intuition behind the methods rather than mathematical theory. Examples will be presented using the Locfit software (an add-on library for S-Plus), which participants will be encouraged to try out. Topics for the lectures are :

1. Local Regression :  Motivation; description of the method; bias and variance properties; model selection.
2. Local Liklihood :  Regressions methods, properties and model selection. Extension to density estimation.
3. Extensions and Applications :  Some of Survival Analysis, Discrimination/Classification, Periodic smoothing and Inferential methods as time permits.