A model of latent class multinomial logit to investigate motorcycle accident injuries

Abstract

The analysis of road traffic accidents will be more complicated with the existence of heterogeneity in the raw data of traffic accident. This study conducts a specific attention to unobserved heterogeneity issues by classifying homogeneous attributes of two different accident data classes. A latent class approach was used to investigate the contributing factors and their influences of motorcycle accident injury outcomes. The data set from 2010 to 2015 consisting 1061 motorcycle accident injuries on Denpasar-Gilimanuk and Denpasar-Singaraja national road networks in Tabanan Regency, Bali were employed as the case study. This study found that male motorists and head on collisions significantly influencing fatal motorcycle injuries. In addition, collisions between motorcycle and the other types of motor vehicles, day time accidents, male motorists at fault, right angle and head on collisions significantly associated with serious motorcycle accident injuries. This result may represent many primary factors which considerably diverge across a traffic accident injury observation. The contributing factors identified in this study were further discussed and some countermeasures for reducing the motorcycle accident injuries were proposed.

Keywords

• accident injuries
• latent class analysis
• motorcycle
• national road network