Compare some estimation methods for zero-inflated Poisson regression models with simulation

Car accidents are an important phenomenon because of their direct relationship to the living conditions of different population centres in cities; it is known that a single accident causes increased human and material losses. For the purpose of researching the topic of predicting the number of accidents, the zero-inflated Poisson distribution was studied. In this thesis, several methods were searched, namely (maximum likelihood estimation, and moments) methods, in order to estimate the zero-inflation parameter of the Poisson regression. In this research, a number of simulation experiments were carried out according to the assumed distribution (Poisson's zero inflation) and the methods for estimating the assumed zero inflation parameter. And for a number of sample sizes (60, 80,100) according to different values of the zero inflation parameters (0.1, 0.2) and the second parameter of the zero-inflated Poisson distribution (1, 2). The comparison between the results of the different simulation experiments was done through the mean square error due to the estimations of each of the two parameters of zero inflation and the second parameter of the zero-inflated Poisson distribution according to each of (estimation method, distribution parameter and sample size).