在線性複迴歸的架構下，Cp是常用的選模統計量。本文透過常態實作模型，使用Royall and Tsou (2003)的強韌概似函數方法，建構Cp的強韌修正項ATp。經由模擬的發現，ATp除了修正Cp的一些缺陷外，在違反常態假設時，表現比Cp另外的修正項RCp(Ronchetti and Staudte, 1994)與RTp(Sommer and Huggins, 1996)好。此外，本文亦將ATp應用在兩個實例上。 Base on multiple linear regression, the statistic, Cp, is usually used to do model selection. In this thesis, we use the robust likelihood technique introduced by Royall and Tsou (2003) to construct a roubust Cp (ATp) under the normal working model. By way of simulations, ATp not only adjusts some defects on Cp, but also is better than RCp (Ronchetti and Staudte, 1994) and RTp (Sommer and Huggins, 1996) which are the other roubust Cp statistics if the normal assumption is wrong. In addition, we use two real examples to demonstate.
