Modeling Censored Wild Boar Data Using Flexible Zero-Inflated Regression

In many practical situations, data exhibit zero-inflation and censoring, which pose challenges for statistical modeling. A notable example is wild boar dispersal distance, which often features a complex structure characterized by an excess of zeros and right-censoring. This study proposes three zero-inflated censored regression models, based on generalized exponential, lognormal, and log-logistic distributions. The likelihood functions for these models are developed, and their bias and mean squared error are assessed through Monte Carlo simulations. To compare the performance of the proposed and existing models, a simulated wild boar dispersal dataset is used – incorporating zero-inflation, right-censoring, and two covariates: age and sex. The findings suggest that the proposed models effectively capture the underlying dispersal patterns, particularly in scenarios where age and sex positively influence dispersal distance. These results address key challenges in drawing reliable inferences from zero-inflated, right-censored wild boar dispersal data.