Apply Appropriate Bias Correction to Crash Data Based on Sample Size and Location Especially for Corridor-Specific Studies
Data Analysis Study on 5-Year Long Crash Data in Richmond, VA Found Lessons Learned on Two Types of Bias Correction
Richmond, Virginia, United States
Pilot Implementation of a Vehicle Occupancy Data Collection Program
Summary Information
Detailed data regarding vehicle occupancy to use in planning decisions is often scarcely available to transportation agencies, due to labor intensity associated with collecting field data. This study developed a methodology for partially automating the extraction of vehicle occupancy data from crash records for the Virginia DOT Richmond District in Virginia, for 2018-2022. The extracted data is then incorporated into online thematic maps showing vehicle occupancies by corridor, block group, jurisdiction (city or county), and district. In addition, the benefits and processing costs of removing potential crash bias were quantified using a statistical approach based on 2019 Richmond District crashes. This study had the ultimate goal of documenting the developed methodology so that it could be replicated in other Virginia DOT districts as necessary, as well as providing insights into lessons learned.
- Apply appropriate bias correction to crash data based on sample size and location especially for Corridor-Specific Studies. Bias correction is necessary for smaller samples (fewer than 1,000 vehicles) to ensure accurate occupancy data. This study suggested two types of bias correction. Type 1 synthesizes what are believed to be missing vehicles from the crash data. Type 1 bias correction, which takes around two hours, is generally sufficient for one-year crash data and yields small improvements, especially in less populous areas. Type 2 measures field occupancies and then uses a regression model from these field estimates to adjust the crash occupancies for a specific corridor. This type offers marginal accuracy improvements but can be time-intensive and sometimes show no benefit, recommended only for sites with higher injury crash occupancies than property damage-only crashes.
- Regularly update the online vehicle occupancy maps. This is crucial to make the most recent data available to the public. In addition, it is important to provide guidance on how the maps should be interpreted.
- Consider using occupancy data for informing planning decisions. This study mentioned that, as occupancies become more widely available, analysts may start to consider how they can be used further to support transportation planning needs.
