227
reads

Yen-Ling Kuo

2009-05-22
12:00:00 - 14:00:00

Analysis of Nonstationary Time Series

201 , Freshman Classroom Building

Variable selection problem is one of the important problems in regression analysis. In this talk, we discuss Mallows' Cp (Mallows, 1973, Technometrics), which was proposed as a way of facilitating comparisons among many alternative subset regressions. First, we will show that the Cp statistics is desvised to balance lack of fit and complexity of model. Next, we describe Mallows' Cp plot with nearly independent predictors and highly correlated predictors, respectively. Finally, we demonstrate how Cp can be used to select the penalty of Ridge Regression estimation (Hoerl and Kennard, 1970, Technometrics), which is often used with highly correlated predictors