Parameter Estimation of Geographically Weighted Compound Correlated Bivariate Poisson Regression Model
Rizki Safarida(a*),Purhadi(a),Sutikno(a)

a)Departement of Statistics,
Institut Teknologi Sepuluh Nopember
Kampus ITS-Sukolilo, Surabaya 60111,Indonesia
*rizkisafarida[at]gmail.com

Abstract

Compound Correlated Bivariate Poisson Regression (CCBPR) is a regression model that combines the Poisson distribution with other distributions that is proposed to overcome overdispersion issue on correlated count data which has high skewness/long tail. Geographically Weighted Compound Correlated Bivariate Poisson Regression (GWCCBPR) is a modified model of CCBPR that consider spatial effect that occur on the data. In this study, GWCCBPR is approached by Generalized Invers Gaussian (GIG) distribution. The aim of this study is to obtain parameter estimation and hypothesis testing for the model GWCCBPR by adding exposure variable. An exposure is included in the model to account for population size difference of the analysis units. The estimation procedure is conducted by using Maximum Likelihood Estimation and BHHH algorithm.

Keywords: BHHH algorithm- Exposure- Geographically Weighted Compound Correlated Bivariate Poisson Regression.

Topic: Mathematics and Statistics