Analytical Study Compared Between Poisson and Poisson Hierarchical Model and Applied in Healthy Field

Abstract

Through this research, We have tried to evaluate the health programs and their effectiveness in improving the health situation through a study of the health institutions reality in Baghdad to identify the main reasons that affect the increase in maternal mortality by using two regression models, "Poisson's Regression Model" and "Hierarchical Poisson's Regression Model". And the study of that indicator (deaths) was through a comparison between the estimation methods of the used models. The "Maximum Likelihood" method was used to estimate the "Poisson's Regression Model"; whereas the "Full Maximum Likelihood" method were used for the "Hierarchical Poisson's Regression Model".  

 The comparison was made through the use of simulation technique, various sample sizes (n= 30, 60, 120) and various frequencies (r= 1000, 5000) for the experiments, The comparison between the estimation methods was built on "Mean Square Errors" method and then to choose the model which most represents the data best. A conclusion was reached, that the "Hierarchical Poissons's Regression Model" - which was estimated by "Full Maximum Likelihood" method with a sample size of (30) – is the most excellent model for representing maternal mortalities data.

  Then this was applied on the real data were obtained from Ministry Of Health. Maternal mortalities were recorded over five years quarterly, Three health institutes in Baghdad were chosen.