Latest News and Events

Using the Kolomgorov-Smirnov test to investigate crash statistics on a rural road

20th April 2017
Using the Kolomgorov-Smirnov test to investigate crash statistics on a rural road
Hugh Dixon, Senior Traffic Engineer at HDS Australia, has recently developed an online tool for using the Kolomgorov-Smirnov test on crash data.  The Kolomgorov-Smirnov test is a statistical test for comparing two data sets and determining whether they have different underlying distributions.  The Austroads (2015, p.65) Guide to Road Safety Part 8 Treatment of Crash Locations mentions it as one of the many statistical tests that can used in before and after analysis to investigate whether changes in crash frequencies are due to a specific treatment or random chance.
  
The tool developed by Hugh Dixon has recently been used to compare the crash distribution on a Victorian rural road linking two rural towns with the distribution on other roads within a 100km radius.  Using the Kolomgorov-Smirnov test indicates whether the dates of the crashes on the road section in question are similar to those on similar roads in the area.  That is, whether there has recently been an increase in number of crashes on this road. The graphical and statistical output of the tool can be viewed at this link. The results indicate that in this case there is no evidence that this road has experienced a ‘crash surge’, and hence we can conclude that the crash history on the road section in question is not out of the ordinary compared to the crash history on surrounding roads.
  
Hugh Dixon will be continuing to look at statistical tests for crash statistics, including further development of this and other tools. He would be happy to discuss statistical analysis of crash data with anyone interested in finding out more. Statistical analysis is an important tool used to ensure that treatments to improve road safety are targeted to locations where they can do the most good, and it is a part of HDS Australia's road safety engineering practice.