Validating crash locations for quantitative spatial analysis
Objective This paper examines the relationship between bicycle collisions and the amount of cycling at the local level.Most previous research has focused on national and city comparisons, little is known about differences within a city (the mesoscale).Does the fact that there are no cyclists injured on a road necessarily mean that the road is safe for cycling?The reason for no cycling casualties can simply be that no one is cycling there.Regardless of the different approaches used in analysing bicycle collision risks, exposure is an essential factor that should be taken into account in risk modelling.14 In the literature, there is no consensus on the best way to measure bicycle exposures.Prevailing exposure measures include time spent on cycling, number of bicycles passing the crossings, number of cycling trips taken and distance travelled calculated from the number of passages or from the mapping of cycling routes by survey respondents based on questionnaires.13 15–18 Recently, Global Positioning System (GPS) data have enabled researchers to gain accurate and precise cycling exposure data, but they are very costly and hence have hardly been used for obtaining cycling flow in the whole territory of a city.As the benefits of cycling can be outweighed by problems of safety and the lack of an adequate infrastructure,5 efforts have been paid to making bicycle use safer.In dealing with the cycling safety issue, one needs to figure out what ‘safety’ means before performing any analysis.
Results The numbers of bicycle collisions went up with the increasing use of bicycles, but the increase in the number of collisions in a given community was less than a linear proportion of the bicycle flow.
The proportion of bicycle usage is much higher in the new towns and New Territories than the dense urban area.22–24 As the local environment such as topography and cycling facilities can vary significantly within a city, it is worth exploring SIN across different neighbourhoods.
The following section will introduce the research methodology.
Following the data description, the way in which cycling exposure is calculated will be presented in details.
Next, negative binomial regression models for modelling bicycle collisions (ie, collisions involving bicycles) will be introduced.