A Simulation Investigation of the Effect of Polluted Observations on Zero-Inflated Poisson Regression

دراسة المحاكاة لتآثير الملاحظات الملوثه عل انحدار بوسون الصفري

Abstract

      Bulk numbers are popular for countable dataries, such as inflated zero regression prototypes. Zero-truncation operations with zero truncation can be effective in elaborating problems with prototypes numbers of zeros, as the behaviour of zeros in the datarie      s shown may elicit difficulties, when they are solved by zero-truncation operations on the databases. The different Poisson threshold prototypes that use are known as the smallest value that gives the shift of the distribution from the normal regression to the zero inflated regression. This research includes simulated experiments according to the difference in the sample size () and the initial values ​​of the Polluted number (), inflation parameter () and (Maximum Likelihood Estimation (MLE), Method of Moment Estimation (MME)) for ZIP with (1000) iterations. The simulation results showed that the inflated zero parameter was affected with (sample size, the distribution parameter value and the initial value of the inflated zero parameter). Other estimation methods can be adopted, including Shrinkage Bayesian and Zero-Inflated Binomial regression models.