Addressing multicollinearity in linear regression: an extended Two Parameter Liu estimator with improved efficiency


ERTAN E.

Communications in Statistics: Simulation and Computation, vol.55, no.7, pp.3049-3064, 2026 (SCI-Expanded, Scopus)

  • Publication Type: Article / Article
  • Volume: 55 Issue: 7
  • Publication Date: 2026
  • Doi Number: 10.1080/03610918.2026.2678503
  • Journal Name: Communications in Statistics: Simulation and Computation
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, MathSciNet, zbMATH, Academic Search Ultimate (EBSCO), Business Source Ultimate (EBSCO), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
  • Page Numbers: pp.3049-3064
  • Keywords: Linear regression, Liu estimator, Mean squared error, Multicollinearity, Ridge estimator, Stein estimator
  • Istanbul University Affiliated: Yes

Abstract

It is well known that the variance of the Ordinary Least Squares (OLS) estimator increases when multicollinearity is present. To overcome this issue, the Two Parameter Ridge Estimator (TRE) has been introduced as an alternative to both the OLS and the Ridge Estimator (RE). This study aims to tackle multicollinearity by introducing a new Two Parameter Liu Estimator (TLE). Moreover, using the Matrix Mean Squared Error (MMSE) and Scaler Mean Squared Error (SMSE) criterions, the superiority of the TLE is examined over OLS, RE, the Liu Estimator (LE) and the TRE. A Monte Carlo simulation is also conducted to evaluate the performances of estimators. The theoretical results are further supported by a numerical example from the literature. Based on both simulation and real world application results, the performance of TLE is better than the other estimators in all cases examined.