Analyzing Correlated Road Accident Count Data Using Zero Truncated Bivariate Poisson Regression Model

Abstract

This paper aims to determine the significant factors which influence two correlated count responses, namely the total number of cars involved in an accident and the total number of fatalities due to that accident, of United Kingdom (UK) road accident count data. The bivariate Poisson (BVP) of two different forms and zero truncated bivariate Poisson regression (ZTBVP) models are considered to analyze UK road accident count data and the best model is selected based on the AIC and BIC values. From the data analysis, it is observed that the ZTBVP model provides the best fit (AIC value: 20563.26) for the UK road accident count data compared to all two variants of the BVP model (AIC value: >20563.26). From the results obtained from ZTBVP model, it is also observed that sex of driver, area, serious severity, and light condition are the significant covariates for the total number of cars involved in an accident while area, fatal severity, serious severity, light condition and year 2021 are the significant covariates for the total number of fatalities due to that accident.