Crash Modification Factor Development: Data Needs and Protocols Raghavan Srinivasan Daniel Carter...
-
Upload
nickolas-armstrong -
Category
Documents
-
view
214 -
download
0
Transcript of Crash Modification Factor Development: Data Needs and Protocols Raghavan Srinivasan Daniel Carter...
Crash Modification Factor Development: Data Needs and
Protocols
Raghavan Srinivasan
Daniel CarterUNC Highway Safety Research Center
Background
• Science of highway safety advancing• Data-driven decisions• Crash modification factors (CMFs) used
more widely– Countermeasure selection– Benefit-cost analysis
• Resources for CMFs expanding
http://www.cmfclearinghouse.org/
http://safety.fhwa.dot.gov/tools/crf/resources/fhwasa10032/
http://safety.fhwa.dot.gov/tools/crf/resources/cmfs/
CMFs in Practice
http://www.cmfclearinghouse.org/collateral/CMF_Protocols.pdf
CMF Protocols
• Need protocols for CMFs– Consistent– Clear in their application– Statistically rigorous– Free of biases– High rating in the CMF Clearinghouse– Acceptable for future editions of the HSM
Audience
• Researchers– Addressing potential biases– Documenting work
• Research sponsors– Writing problem statements– Evaluating contractor work
Scope
• Crash-based• Infrastructure-related countermeasures
(signs, signals, markings, barriers, etc)
Background Work
• Literature review across many study methods
• Web survey and phone interviews with CMF researchers
How Prescriptive?• Protocols could be very prescriptive• Specify study methodology (e.g., empirical
Bayes)– May not be appropriate– May not be possible– May oppress use of more innovative methods
• Specify statistical confidence level– Depends on sample size and data variability– Funding, time period of data, availability of sites
What is the Intent?
• If not prescriptive, then what?• What is the intent?
– Consistent– Clear in their application– Statistically rigorous– Free of biases– Acceptable for HSM and high rating in the CMF
Clearinghouse
Need to know how CMF was developed
Need to know how potential biases were addressed
Documentation
• General documentation (for determining applicability)
• Biases documentation (for determining quality)
General Documentation
• CMF and Countermeasure– Detailed countermeasure description: “Install 3-
strand high-tension cable median barrier”– Prior condition: “Depressed grass median
without barrier”– The Crash Modification Factor or Crash
Modification Function– Measure of precision of the CMF
General Documentation
• Site characteristics– Roadway type: 4 to 6 lane freeway– Geographic area: Rural– Traffic volume range: 20,000 to 40,000 AADT– Other site characteristics
General Documentation
• Crash characteristics– Type: Cross-median crashes– Severity: Fatal and injury crashes– Other crash characteristics
General Documentation
• Study details– Sample size: 210 miles– Years: 6 years (3 before, 3 after)– Selection criteria: High incidence of target
crashes– Other study details
Biases Documentation
• Cross-sectional studies• Before-after studies• Issues that affect all studies
Potential Cross-Sectional Biases
• Confounders• Heterogeneity / omitted
variable• State-to-state differences • Appropriate functional form• Correlation or collinearity • Overfitting • Small sample mean and
size• Aggregation, averaging, or
incompleteness in data
• Temporal and spatial correlation
• Endogenous independent variables
• Misspecification of structure of systematic variation and residual terms
• Correlation between crash types and injury severities
Potential Before-After Biases
• Regression-to-the-mean• Changes in traffic volumes• Trends• Changes in crash reporting• State-to-state differences if using multiple states• Suitability of comparison or reference groups• Other biases for before-after studies
Issues That Affect All Studies
• Data quality (abrupt increases/decreases, suspicious trends)– Crashes– AADT
• Crash type and severity• Crash modification factor vs. function
Overall Objective
Better CMF
Documentation
Better Knowledge of CMF Applicability
Better Knowledge of CMF Quality
Better CMF Studies