A POISSON REGRESSION MODEL OF FATAL ROAD ACCIDENTS ON SELECTED ROAD NETWORKS IN THE BRONG AHAFO REGION, GHANA

Abstract

Models for predicting fatal accidents on some selected road networks in the Brong Ahafo Region in Ghana were developed using Poisson regression model. The objectives of the study included both identificati on of factors contributing to the incidence of fatal accidents and the development of models to predict such occurrences on the selected road networks. The data on records of incidence of accidents in general in the years 2008 - 2010 , generated by the Motor Traffic and Transport Unit (MTTU ) of the Ghana Police Service , Brong Ahafo Region. Other information required for the study were taken or measured from the road network s selected for the study. The research hypothesis states that, the occurrence of a fatal accident depends on at least a risk factor . The data was analysed using the SAS package version 9 .2 . 2 and result that the most significant variables that contribute to the occurrence of fatal accidents include traffic volume and section length out of a total of seven variables identified besides the driver and vehicle characteristics which were not considered in this study. From the analysis, the Poisson regression mode l fitted the data well compared to other distributions such as the Gamm a, Negative binomial, zero - inflated Poisson and the zero inflated Negative binomial models. Finally, the models developed were validated using the graphical approach of the residual analysis