Almost unbiased ridge estimator in the zero-inated Poisson regression model

Abstract

The zero-inflated Poisson regression (ZIP) model is a very popular model for count data that have extra zeros. In some situations, the count data are correlated and so multicollinearity exists among the explanatory variables. Thus, the traditional maximum likelihood estimator (MLE) becomes not a reliable estimator because the mean squared error (MSE) becomes inflated. The ridge estimator (RE) is used to overcome this problem. In this work, an almost unbiased ridge estimator for the ZIP model (AUZIPRE) is proposed to tackle the multicollinearity problem in count data. We investigate the behavior of the proposed estimator using a simulation study. Using the MSE measure, the results of the proposed estimator are compared with those of the RE and the MLE. Furthermore, we apply the proposed estimator on a real dataset. The results show that the performance of AUZIPRE outperforms for that of the RE and the MLE in the existing of the multicollinearity among the count data in the ZIP model.