Mihaela Covrig (mihaelacovrig@gmail.com)
Iulian Mircea (mirceaiulian91@yahoo.com)
Bucharest University of Economic Studies
Gheorghiţă Zbăganu (georghitazbaganu@yahoo.com)
University of Bucharest
Alexandru Coşer (alexandru.coser@gmail.com)
Vodafone Romania
Alexandru Tindeche (tindeche_alex@yahoo.com)
BCR Asigurări de Viață (BCR Life Insurance)
Abstract
This paper aims to approach the estimation of generalized linear models (GLM) on the basis of the glm routine package in R. Particularly, regression models will be analyzed for those cases in which the explained variable follows a Poisson or a Negative Binomial distribution. The paper will briefly present the GLM methodology for count data, while the practical part will revolve around estimating and comparing models in which the response variable shows the number of claims in a portfolio of automobile insurance policies.
Key words: GLM, count data, insurance, Poisson regression, Negative Binomial Regression, R
JEL: C250
