101
reads

Chuan-Fa Tang

2009-08-28
12:30:00 - 14:30:00

Adaptive Linear Regression Selection

308 , Mathematics Research Center Building (ori. New Math. Bldg.)

Variable selection problem is one of the most important problems in regression analysis. In this talk, we first describe several model selection criterions such as Akaike’s information criterion (AIC) (Akaike 1973), Mallows’s Cp (Mallows 1973), and Bayesian informationcriterion (BIC) (Schwarz 1978). Under normal error, those three criteria can be written as final prediction error which is the sum of residual sum squares and penalty. Here the penalty is a linear function of number of regressors. Then evaluate the performance of those criteria using several data sets in Shen and Ye (2002). Finally, we address the need of new adaptive model selection criterion.