Report number | RA-2006-87 |

Title | Calculation of the effect of a traffic safety measure on the average speed and the V85 |

Subtitle | |

Authors | Erik Nuyts |

Published by | Policy Research Centre for Traffic Safety 2002-2006 |

Number of pages | 36 |

Date | 22/05/2006 |

ISBN | |

Document language | Dutch |

Partner(s) | PHL |

Work package | Other: Infrastructure and space |

Summary | It is not always possible to measure directly the effect of a traffic safety measure on the number of accidents. In a limited time window, the number of accidents is not always sufficiently large to allow statistical testing. Where traffic safety is linked with traffic the effect of the measure could be approximated by measuring the effect on the traffic speed.
In Flanders, when the effect of a measure on traffic speed is calculated, most often there lack a correction for the general trend of traffic safety. Also statistical tests to see if the effect is significant are hardly used. In this report, the necessary formulas to do so are shown. Since not all members of the target group of the method have the disposal of advanced statistical packages, only formulas are used that can be calculated with regular spreadsheets like Microsoft Excel.
The formulas allow estimating the effect on the mean speed and on the V85, with and without a comparison group. As an extension of what is found in literature, calculations in this report allow the use of more than one location as a comparison group.
From the formulas, it follows that the lack of a comparison group results in an underestimation of the standaard error of the estimate. Hence, tests without an comparison group find easily significant results, that would not be significant if a comparison group was used.
After calculating the effect of a measure on several locations, one often wants to have some general value of the effect of the measure. For this purpose, two formulas are presented, based on the theory of meta-analysis. One method (fixed effects model) allows to test if the measures were successful at the locations where the measures were taken. A second method (random effects method) allows estimating if the same measure would be successful in new locations. The second method is much more conservative than the first method. Hence, it is possible to find a significant effect for the locations where the traffic safety measure has been performed, without having a significant effect for extrapolation to other locations.
All methods are explained with an example of variable speed message signs in Antwerp. |

Download | RA-2006-87.pdf |