Modelling the number of traffic accident using negative binomial regression spline

Abstract

Data from World Health Organization (WHO) in world showed every year more than 1.25 million victims die from traffics accidents. East Java Province had the highest number of traffic accident cases in Indonesia. Then to predict the number of accidents in East Java against predictor variables, it is necessary to build a model. Traffic accident could make victims die, serious injury, and light injury. In this study, the researcher estimate the parametric and nonparametric regression negative binomial model based on the truncated spline estimator. The best model is determined based on Maximum Likelihood Cross-Validation (MLCV). The nonparametric regression model with negative binomial approach with the best-truncated spline estimator is obtained from the combination of knots (2, 1, 2, 3) using the MLCV method. The comparison of the deviance values between parametric and nonparametric regression in this study showed the deviance value nonparametric model less then deviance value parametric model. Deviance values showed the binomial negative using nonparametric regression model approach based on truncated spline estimator is better than the negative binomial using parametric regression model approach. The analysis showed factor traffic accident cause driver is sleepy had the highest influence on the number of traffic accidents cases in East Java Province.