Sufficient jackknife-after-bootstrap method for detection of influential observations in linear regression models


Beyaztas U. , Alin A.

STATISTICAL PAPERS, vol.55, no.4, pp.1001-1018, 2014 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 55 Issue: 4
  • Publication Date: 2014
  • Doi Number: 10.1007/s00362-013-0548-4
  • Title of Journal : STATISTICAL PAPERS
  • Page Numbers: pp.1001-1018
  • Keywords: Sufficient bootstrap, Jacknife, Bootstrap, Influential observation, Regression diagnostics

Abstract

In this study, we adapt sufficient bootstrap into the jackknife-after-bootstrap (JaB) algorithm. The performances of the sufficient and conventional JaB methods have been compared for detecting influential observations in linear regression. Comparison is based on two real-world examples and an extensive designed simulation study. Design includes different sample sizes and various modeling scenarios. The results reveal that proposed method is a good competitor for conventional JaB method with less standard error and amount of computation.