This paper illustrates the use of a modern statistical modelling approach to road casualty data and the analysis of performance indicators. The generalised linear mixed model (GLMM) approach is applied to motorcycle casualty counts using the number of registered motorcycles in each authority rather than per capita resident population data. Motorcycle casualties reported between 1999 and 2002, corresponding to Value Performance Indicator 99 data years 2000-2001, are used. It was assumed that the casualty count follows a Poisson distribution. Random effects were modelled by including an extra variable in the regression model used to model the distribution. The results for the killed and seriously injured casualty rate are presented for different local authorities in the UK. It is noted that counties with rural areas adjacent to conurbations had higher motorcycle accident rates. The slight injury rate was highest for London and lowest for rural counties.
Abstract