A novel approach for zero-inflated count regression model: Zero-inflated Poisson generalized-Lindley linear model with applications

Count regression models are important statistical tools to model the discrete dependent variable with known covariates. When the dependent variable exhibits over-dispersion and inflation at zero point, the zero-inflated negative-binomial regression model is used. The presented paper offers a new model as an alternative to the zero-inflated negative-binomial regression model. To do this, Poisson generalized-Lindley distribution is re-parametrized and its parameter estimation problem is discussed via maximum likelihood estimation method. The proposed model is called as zero-inflated Poisson generalized Lindley regression model. The results regarding the efficiency of parameter estimation of the proposed model are evaluated with two simulation studies. To evaluate the success of the proposed model in the case of zero inflation, two datasets are analyzed. According to the results obtained, the proposed model gives better results than the negative-binomial regression model both in case of over-dispersion and in the case of zero inflation.

Automated summary

Count regression models are important for modeling discrete dependent variables with known covariates. This study introduces a new model, the zero-inflated Poisson generalized Lindley regression model, to replace the zero-inflated negative-binomial regression model. The Poisson generalized-Lindley distribution is re-parametrized and its parameter estimation problem is discussed using maximum likelihood estimation. The efficiency of parameter estimation of the proposed model is evaluated through simulation studies and the success of the model in handling zero inflation is tested using two datasets. The proposed model outperforms the negative-binomial regression model in cases of over-dispersion and zero inflation.