Comparison Between Tree regression (TR), and Negative binomial regression (NBR) by Using Simulation.

Abstract

In this paper, the process of comparison between the tree regression model and the negative binomial regression. As these models included two types of statistical methods represented by the first type "non parameter statistic" which is the tree regression that aims to divide the data set into subgroups, and the second type is the "parameter statistic" of negative binomial regression, which is usually used when dealing with medical data, especially when dealing with large sample sizes. Comparison of these methods according to the average mean squares error (MSE) and using the simulation of the experiment and taking different sample sizes where the results of simulation showed that the tree regression is best when the value of variance is large (5) and for all sample sizes model negative binomial regression when variance values (0.01, 0.5, 1) for all sample sizes, this method is superior to tree regression only when we take medium sample sizes.