|
Activity Number:
|
117
- Need for New Statistical Methods to Save Lives on Highways
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 31, 2017 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Transportation Statistics Interest Group
|
|
Abstract #324735
|
|
|
Title:
|
Need for New Statistical Methods to Save Lives on Highways
|
|
Author(s):
|
Eric Donnell* and R. J. Porter* and Larry Cook* and Raghavan Srinivasan* and Joseph Kufera*
|
|
Companies:
|
The Pennsylvania State University and VHB and University of Utah and University of North Carolina - Highway Safety Research Center and University of Maryland
|
|
Keywords:
|
Traffic Safety ;
Causal Inference ;
Regression Trees ;
Underreporting ;
CODES Data
|
|
Abstract:
|
The transportation engineering community is undergoing transformative change by integrating quantitative methods into the project development process. This session will compare current statistical analysis methods and data sources used in traffic safety research with alternative methods and data sources. The intent is to identify opportunities to further understand the relationship between roadway safety performance and the factors that affect traffic crash occurrence and severity outcomes. In this session, causal inference methods will be compared to observational before-after methods to develop safety effect estimates of centerline and edgeline rumble strips. Regression trees and random forests will be compared to count regression methods to predict crash frequencies on freeways. Traffic safety performance estimates using the Crash Outcomes Data Evaluation System (CODES) will be discussed, with a focus on opportunities to link hospital and crash data to understand the relationship between crashes and site-specific contributing factors. Methods to account for underreporting in crash frequency models are also described.
|
Authors who are presenting talks have a * after their name.
