On Simultaneous Considerations of Variable Selection and Detection of Influential Cases

  • Ahn, Byoung-Jin (Department of Applied Statistics, Kon-kuk University, Seoul 133) ;
  • Park, Sung-Hyun (Department of Computer Science and Statistics, Seoul National University, Seoul 151)
  • Published : 1987.06.01

Abstract

The value of statistics used for variable selection criteria can be reduced remarkably by excluding only a few influential cases. Furthermore, different subsets of regressors change leverage and influence patterns for the same response variable. Based on these motivations, this paper suggests a procedure which considers variable selection and detection of influential cases simultaneously.

Keywords