Factor Association with Multiple Correspondence Analysis in Vehicle–Pedestrian Crashes

Abstract

In the United States, about 14% of total crash fatalities are pedestrian related. In 2012, 4,743 pedestrians were killed, and 76,000 pedestrians were injured in vehicle–pedestrian crashes in the United States. Vehicle– pedestrian crashes have become a key concern in Louisiana as a result of the high percentage of fatalities there in recent years. In 2012, pedestrians accounted for 17% of total crash fatalities in the state. This study used multiple correspondence analysis (MCA), an exploratory data analysis method used to detect and represent underlying structures in a categorical data set, to analyze 8 years (2004 to 2011) of vehicle–pedestrian crashes in Louisiana. Pedestrian crash data are best represented as transactions of multiple categorical variables, so the use of MCA was a unique choice to determine the relationship of the variables and their significance. The findings indicated several nontrivial focus groups (e.g., drivers with high-occupancy vehicles, female drivers in bad weather conditions, and drivers distracted by mobile phone use). The associated geometric factors were hillcrest roadways, dip or hump aligned roadways, roadways with multiple lanes, and roadways with no lighting at night. Male drivers were seen to be relatively susceptible to severe and moderate injury crashes. Fatal pedestrian crashes were correlated to two-lane roadways with no lighting at night. The MCA method helped measure significant contributing factors and degrees of association between the factors through the analysis of the systematic patterns of variation with categorical data sets of pedestrian crashes. The findings from this study will help transportation professionals improve countermeasure selection strategies.