The aim of this research was to assess the potential of innovative forms of statistical analysis to explore the STATS19 database of records road accidents in Great Britain. It contains many variables which in the past have tended to be analysed in relatively small groups. The report investigates a group of techniques called cluster analysis which aim to identify homogeneous groups of accidents of interest for further analysis. Three forms of cluster analysis were investigated: cross-tabulation, two-step analysis and CHAID analysis. CHAID had the greatest potential; the results could be interpreted objectively and there was little scope for researchers to introduce bias. The clusters are produced by a transparent process that can be interpreted without any formal statistical training. There is also a clear measure of which clusters are of the greatest interest, in this case accident severity. The second part of the report investigated the effect of the sun's position on the incidence of accidents. It investigates the relative propensity for drivers to be involved in an accident according to the sun's position relative to the driver at the time of the accident, and finds clear evidence of an influence. (Author/publisher)
Abstract