2018Università del SalentoOpen access

Comparison of regression models under multi-collinearity

H. Bagya Lakshmi, Michele Gallo, Rangasami M. Srinivasan

Open full text 4 citations

Abstract

Multicollinearity is a major problem in linear regression analysis and several methods exists in the literature to deal with the same. Ridge regression is one of the most popular methods to overcome the problem followed by Generalized Ridge Regression (GRR) and Directed Ridge Regression (DRR). However, there exist many computational issues in using the above methods. Partial Ridge Regression (PRR) method is a computationally viable approach by selectively adjusting the ridge constants using the cutoff criteria. In this paper, the performance of the Partial Ridge Regression approach has been evaluated through a simulation study based on the mean squared error (MSE) criterion. Comparing with other methods of ridge regression, the study indicates that the Partial ridge regression by cutoff criteria performs better than the existing methods.

About this research paper

What this paper is about

Multicollinearity is a major problem in linear regression analysis and several methods exists in the literature to deal with the same. Ridge regression is one of the most popular methods to overcome the problem followed by Generalized Ridge Regression (GRR) and Directed Ridge Regression (DRR). However, there exist many computational issues in using the above methods. Partial Ridge Regression (PRR) method is a computationally viable approach by selectively adjusting the ridge constants using the cutoff criteria. In this paper, the performance of the Partial Ridge Regression approach has been evaluated through a simulation study based on the mean squared error (MSE) criterion. Comparing with other methods of ridge regression, the study indicates that the Partial ridge regression by cutoff criteria performs better than the existing methods.

Why it matters

OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

Multicollinearity is a major problem in linear regression analysis and several methods exists in the literature to deal with the same. Ridge regression is one of the most popular methods to overcome the problem followed by Generalized Ridge Regression (GRR) and Directed Ridge Regression (DRR). However, there exist many computational issues in using the above methods. Partial Ridge Regression (PRR) method is a computationally viable approach by selectively adjusting the ridge constants using the cutoff criteria. In this paper, the performance of the Partial Ridge Regression approach has been evaluated through a simulation study based on the mean squared error (MSE) criterion. Comparing with other methods of ridge regression, the study indicates that the Partial ridge regression by cutoff criteria performs better than the existing methods.

Key concepts: Multicollinearity, Ridge, Regression, Elastic net regularization, Regression analysis, Collinearity, Statistics, Cutoff

Related papers

Back to paper searchBrowse research topicsOriginal source
Comparison of regression models under multi-collinearity — Research Paper | ScholarLens