Modified two-parameter ridge estimators with enhanced performance under multicollinearity in linear regression: simulation and applications
Year: 2026
Type: article
Abstract: Multicollinearity among predictor variables remains a persistent challenge in regression analysis, often rendering ordinary least squares (OLS) estimates unreliable and forcing researchers to drop correlated predictors thereby risking a significant loss of valuable information. Ridge regression offers a viable alternative by retaining all predictors and stabilizing coefficient estimates; however, no single ridge estimator has been found to perform optimally across all multicollinearity conditions. To address this gap, the present study introduces improved two-parameter ridge estimators (ITPREs) that explicitly incorporate the variance of the error term into the estimation process. The proposed estimators are designed to handle multicollinearity more effectively than OLS and a range of existing ridge-type methods. Their performance is evaluated using mean square error (MSE) as the primary criterion. Extensive simulation results demonstrate that the proposed TPREs often yield the smallest estimated MSEs, particularly under high levels of multicollinearity, outperforming both OLS and conventional ridge estimators. Furthermore, applications to real-world datasets illustrate the practical utility and reasonable performance of the new estimators, making them a valuable addition to the toolkit for multicollinearity mitigation.
Source: Kuwait Journal of Science
Authors: Nadeem Akhtar, Muteb Faraj Alharti
Institutions: Government of Khyber Pakhtunkhwa, Taif University
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,100
Funder: Taif University
OpenAlex ID: W7235763240
