A New Ridge Type Estimator in the Logistic Regression Model with Correlated Regressors

Abstract

The maximum likelihood (ML) technique is always one of the most widely employed to estimate model parameters in logistic regression models. However, due to the problem of multicollinearity, unstable parameter estimates, and inaccurate variance which affects confidence intervals and hypothesis tests can be achieved. A new two-parameter biased estimator is proposed in this paper to handle multicollinearity in binary logistic regression models. The proposed estimator's properties were determined, and five (5) different types of biasing parameter k (generalized, maximum, median, mid-range, and arithmetic mean) were applied in this work. The necessary and sufficient criteria for the new two-parameter biased estimators to outperform the existing estimators is considered. In addition, Monte Carlo simulation studies are carried out to compare the performance of the proposed biased estimator. Finally, a numerical example is provided to support the theoretical and simulations findings.