A new correction approach for information criteria to detect outliers in regression modeling

Dünder E.

COMMUNICATIONS IN STATISTICS-THEORY AND METHODS, vol.50, no.10, pp.2451-2465, 2021 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 50 Issue: 10
  • Publication Date: 2021
  • Doi Number: 10.1080/03610926.2020.1792497
  • Journal Name: COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Business Source Elite, Business Source Premier, CAB Abstracts, Compendex, Veterinary Science Database, zbMATH, Civil Engineering Abstracts
  • Page Numbers: pp.2451-2465
  • Keywords: Corrected information criteria, lasso, regression modeling, outlier detection, ROBUST REGRESSION, SELECTION, ALGORITHMS
  • Ondokuz Mayıs University Affiliated: Yes

Abstract

The outliers cause wrong prediction and estimation results in regression models. Therefore, it is important to identify the outliers correctly in the context of regression analysis. Information criteria can be used to perform this task with corrections but these corrected versions of criteria require some subjective parameters. In this article, an objective correction approach is proposed within the information criteria to perform outlier detection in regression modeling. The evaluations are performed on lasso regression. The numerical examples demonstrate that the proposed correction successfully achieves the outlier detection task in regression models without requiring any subjective correction parameter.