Count data regression models and an application

Regression analysis is used to measure the relationship between a dependent variable and one or more explanatory variables. When the dependent variables consist of discrete non-negative values, the use of counting data models is recommended. Poisson regression which is one of the most common models used in count data analysis. In the Poisson regression analysis, mean and variance are assumed to be equal, but when this equality is not provided, underdispersion or overdispersion occurs. For these cases, negative binomial and generalized Poisson regression models have been developed. In this study, generalized linear models were examined, Poisson regression model, negative binomial regression model and zero-inflated models were introduced. The data taken from Turkish Statistical Institute, Household Labour Force Survey for 2016 were modeled using negative binomial regression and generalized Poisson regression analysis. The "working" variable which contains knowledge of persons' how many hour working in a week is dependent variable. Sector, age, gender, education, occupation, registration, continuity are taken as explanatory variables. The created regression models are compared with model selection criteria. The best model for the dataset applied to data as negative binomial regression model. In practice, SPSS, SAS and R programs were used to obtain results.