Abstract:When overdispersion and zero-inflated feature are both concerned, Zero-inflated negative binomial (ZINB) model tends to be more appropriate than the popular Zero-inflated Poisson (ZIP) model most used. Moreover, when some variables that model the mean response and the probability of zero-excess have missing data, ignoring the missing or using inappropriate methods is likely to produce questionable estimates. To remedy this problem, we propose a non-parametric and parametric inverse probability weighting (IPW) estimators to estimate parameters of the ZINB models. In particular, the generalized additive models (GAMs), the indicator or kernel smoothing functions are considered for the non-parametric approaches. Meanwhile, logistic model is used to estimate selection probability in parametric method. The performance of our proposed methods are investigated by conducting several simulation studies and exploring a real example analysis. We aim to develop and compare several IPW estimators in order to recommend the most accurate approaches. Numerical illustrations support that the semiparametric IPW methods which were based on the kernel smoothing or indicator function weights tend to produce the best results, because they tend to outperform the GAMs and parametric IPW estimates in terms of efficiency. Hence we recommend that the first approach to be used in the analysis for the ZINB model should be the semiparametric IPW based on the kernel smoothing or indicator function weights. The GAM and parametric weighting estimators are also quite good, as long as their related estimated weights are correct.

Zero-Inflated Generalized Poisson Regression Model and Its Application to Insurance Ratemaking

Simulating Posterior Distributions for Zero-Inflated Automobile Insurance Data

A comparative analysis of several multivariate zero-inflated and zero-modified models with applications in insurance

Risk factor selection in rate making: EM adaptive LASSO for zero-inflated poisson regression models.

Zero to k Inflated Poisson Regression Models with Applications

The Zero-Inflated Poisson - Probit regression model: a new model for count data

Zero-inflacted Poisson regression, with an application to defects in manufacturing

Logistic and ZIP Regression Model for Forest Fire Data

A Zero-Inflated Ordered Probit Model, with an Application to Modelling Tobacco Consumption

ASYMPTOTIC PROPERTIES IN THE PROBIT-ZERO-INFLATED BINOMIAL REGRESSION MODEL

Geographically weighted bivariate zero inflated generalized Poisson regression model and its application

Predicting the multivariate zero‐inflated counts: A novel model averaging method under Pearson loss

Parameter Estimations on Zero-Inflated Negative Binomial Model with Incomplete Data

The Compound Poisson Random Variable'S Approximation to the Individual Risk Model

Mapping QTL controlling count traits with excess zeros and ones using a zero‐and‐one‐inflated generalized Poisson regression model

Dynamic online prediction model and its application to automobile claim frequency data

An application of Zero-One Inflated Beta regression models for predicting health insurance reimbursement

Tweedie Gradient Boosting for Extremely Unbalanced Zero-inflated Data

A New Zero–Inflated Regression Model with Applications to Australian Health Survey and Biochemistry Graduate Students Data

EM Estimation for the Poisson-Inverse Gamma Regression Model with Varying Dispersion: An Application to Insurance Ratemaking

Zero truncated Poisson model: an alternative approach for analyzing count data with excess zeros