An evaluation of some ridge parameter estimators in the binary logistic regression model via two new additional criteria

OpenAlex · İngilizce

Multicollinearity is a common issue in regression analysis, often addressed through methods such as Ridge Regression (RR) and Principal Component Regression (PCR), with RR being widely employed in the literature. This study analyzes the impact of multicollinearity on the Maximum Likelihood (ML) estimator within the framework of the binary logistic regression model. To this end, a Monte Carlo simulation and an empirical data analysis were conducted to evaluate the performances of 15 distinct ridge parameter estimators. The objective is to determine the ridge parameter estimator that minimizes the Mean Squared Error (MSE) across 175 different scenarios, considering variations in sample size, the degree of multicollinearity, and the number of independent variables. In contrast to previous studies, this paper not only considers MSE as a performance criterion but also examines the adherence of ridge parameter estimates to normality and their range within [0-10], emphasizing the importance of a non-skewed distribution of ridge parameter estimators in the binary logistic ridge regression model.