Conway-Maxwell-Poisson regression model

The regression model is used to determine the relationship between dependent variable and one or more independent variables. Counting/discrete dependent variable is encountered in many studies. In this case, instead of the classical regression model, It is more appropriate to use generalized linear models that including many distributions Poisson, negative binomial et al. Although the Poisson regression model is frequently used in the analysis of counting data, the assumption of the equality of mean and variance is not provided for many experimental data. It is more appropriate to analyze with alternative regression models in case of over-under dispersion. Generally, the negative binomial regression model is widely used for overdispersed data. In this thesis, the Conway-Maxwell-Poisson Regression model which is suitable for analysis of both over-under dispersed data is presented in detail. Beside Conway-Maxwell-Poisson distribution information, parameter estimation, significance and interpretation are given. Poisson, negative binomial and COM-Poisson regression models were used to determine the effect of socio-demographic factors on the number of visits Giresun Community Mental Health Center (TRSM) between 2011 and 2014. Then, three regression models applied to model school type, gender, and age group factors with the number of social media accounts of the teachers. All possible subset regression approach was used as a variable selection method for both data sets. The smallest AIC-valued model was selected as the best model. COM-Poisson regression was the best model in both datasets. The analyses were performed by using the "MASS "and "COMPoissonReg" packages and the "glm" function on the R program.