Enhancing Statistical Methodologies For Highway Safety Research – Impetus From FHWA

INTRODUCTION

Background on the Development of Crash Modification Factors (DCMF) Program

The Federal Highway Administration (FHWA) DCMF program was established in 2012 to address highway safety research needs for evaluating new and innovative safety strategies (improvements) by developing reliable quantitative estimates of their effectiveness in reducing crashes.

The ultimate goal of the DCMF program is to save lives by identifying new safety strategies that effectively reduce crashes and promote them for nationwide installation by providing measures of their safety effectiveness and benefit to cost ratios (B/C) through research. State transportation departments and other transportation agencies need to have objective measures for safety effectiveness and B/C ratios before investing in new strategies for statewide safety improvements. There are 38 State transportation departments that provide technical feedback on safety improvements to the DCMF program and implement new safety improvements to facilitate evaluations. These States are members of the Evaluation of Low Cost Safety Improvements Pooled Fund Study, which functions under the DCMF program.

Statistical methodologies are heavily used for all studies performed under the DCMF program, but these methodologies have been borrowed from various statistical fields and have limitations in capability and applicability when used for highway safety research. Accordingly, a secondary goal of DCMF program is to advance highway safety and related research by establishing a sound foundation for the development of highway transportation specific statistical methodologies in cooperation with the American Statistical Association (ASA) and other statistician communities.

This white paper may be considered as the first brick for laying the foundation for future highway transportation specific statistical methodologies and/or a future statistical field customized for research and application in various highway transportation areas.

Project Background

The goals of the project on which this white paper is based were as follows:

• Identify new statistical methodologies.
• Improve and advance the current methodologies used in the development of Safety Performance Functions (SPFs) and Crash Modification Factors (CMFs).
• Suggest an initial set of reliable and time/labor efficient methodologies that can be used by the FHWA and State transportation departments and other transportation agencies for SPF and CMF development.

Working toward these goals, the project team, in consultation with FHWA, planned a two-day technical experts meeting to bring together researchers from the road safety, statistics, and other statistics-related fields such as epidemiology, biostatistics, and agent based modeling that have methodologies relevant to highway safety research applications. The meeting was arranged to work with attendees in identifying applicable advanced methodologies and/or improve existing methodologies for the development of SPFs and CMFs.

Prior to the meeting, introductory/educational material for highway safety improvement evaluations and sample data were provided to the invited statisticians and other experts. This advance information was intended to prepare the experts to share their assessments of the current practice in highway safety improvement evaluations, identify opportunities and issues, and make initial suggestions at the end of the meeting on ways to improve the existing methodologies.

This white paper is the final product of this project and has been developed based on the outcome of the meeting to record the expert assessments of current practices for highway safety improvement evaluations, opportunities and issues, the suggestions that were made, and possible future steps for advancing the safety evaluation methodologies. Since the paper is specific to what was presented and discussed at the meeting, it is of necessity not an exhaustive assessment of needs related to the estimation of SPFs and CMFs. Rather, the scope is limited to an initial set of suggestions for improving estimation methodologies. In some cases the suggestions pertain to methodologies that have been investigated by highway safety researchers. Reporting on these methodologies as an outcome of a meeting of statistical experts provides support for continued and expanded research in those areas.

It is expected that a key benefit of the project will be the initiation of dialog, and collaboration between, highway safety researchers, statisticians, and other experts to work together to continue to find new, and/or improve current methodologies for the development of SPFs and CMFs.

Selection and Preparation of Technical Experts

The project team, in coordination with FHWA, identified and invited statistical and other technical experts with expertise in methodologies that may be applicable to highway safety improvement evaluations or the development of prediction models. Selections were based on suggestions from both the FHWA and project team (most notably, Dr. David Banks in his capacity as chair of the ASA Transportation Interest Group, who also served as one of the experts). The selection of technical experts was guided by several considerations, including the following:

• They should have extensive experience in statistics, including a preference for a Ph.D. in statistics.
• They will not have been directly involved in the evaluation of road safety engineering treatments related to infrastructure.
• They should have expertise/experience in some sort of treatment/program evaluation that requires the statistical analysis/modeling of time series and/or cross-sectional data.
• They should have an established track record that qualifies them as experts in their field.

The project team also sought to have the sample of technical experts drawn from the following categories:

• Category 1: Experts with no experience in road safety data analysis who have related expertise in some sort of treatment/program evaluation that requires the statistical analysis/modeling of time series and/or cross-sectional data.
• Category 2: Experts in transportation engineering and/or statistics with extensive seminal research in statistical analysis of crash data but who have not been involved directly in CMF or SPF development.
• Category 3: Experts who have worked in road safety evaluations related to driver and enforcement programs.

The chosen experts included five from category 1, two from category 2, and one from category 3. Appendix A lists the invited technical experts, along with their affiliations and areas of expertise.

Experts in the development of SPFs and CMFs were intentionally excluded, given the objectives of the project to generate ideas from statisticians in other areas. However, several members of the project team with considerable experience in SPF and CMF estimation were present at the meeting and ensured that the discussion focused on the task at hand.

The project team communicated with the selected experts to share material and data for highway safety improvement evaluations in advance of the technical experts meeting for the purpose of preparing them for the meeting. In the preparation period, the project team provided further information and assistance, and answered questions on the provided material and data. The materials provided included the following:

• A document providing an overview of SPFs and CMFs and issues in their development.
• A Guide to Developing Quality Crash Modification Factors, FHWA-SA-10-032.
• How to Choose Between Calibrating SPFs from the HSM and Developing Jurisdiction-Specific SPFs, a report from Project TPF-5(255) of the Transportation Pooled Fund Program.
• Safety Evaluation of Improved Curve Delineation, FHWA-HRT-09-045, a report for a completed Empirical Bayes (EB) before-after evaluation.
• Safety Evaluation of Lane and Shoulder Width Combinations on Rural, Two-Lane, Undivided Roads, FHWA-HRT-09-031, a report for a completed case-control evaluation.

Technical Experts Meeting and Follow Up

The technical experts meeting was held for two days in December 2013 in Raleigh, NC, at the National Institute of Statistical Sciences. Following a review of the background material shared, each invited expert presented on his or her work and how it could relate to similar problems in the development of SPFs and CMFs. This was followed by extensive discussion and brainstorming. The products produced from the meeting included the following:

• Meeting notes.
• A list of concerns and issues.
• A list of opportunities for improving transportation safety research.
• Initial suggestions for new methodologies or methodology improvements.
• Possible follow ups and next steps.
• A draft marketing and communications plan.

The final project task is the development of the white paper and a final marketing and communication plan. This document constitutes the white paper, which is organized as follows:

• Chapter 1 provides background information on crash modification factors and safety performance functions, mainly for the benefit of statisticians who have not worked in the area of road safety analysis.
• This section provides the backdrop for chapter 2, in which opportunities for advancing methodologies for estimating CMFs and SPFs are identified and discussed.
• Finally, chapter 3 provides considerations and future steps, with a view to ensuring that the opportunities identified for enhancing the SPF and CMF estimation methodologies will be seized by researchers.
• Appendices A-C provide details on the technical experts, an overview of commonly applied methods in estimating CMFs, and an assessment of issues in current highway safety improvement evaluations, respectively.
• Appendix D provides summaries for a small sample of recent road safety research that applied some of the innovative statistical techniques identified at the technical experts meeting. The intent of this appendix is to illustrate that the application of these advanced techniques is topical and is not beyond the capabilities of road safety researchers and, in so doing, to stimulate others to pursue these techniques in their research to develop CMFs and SPFs.
• Appendix E provides a summary of additional ideas for discussion generated by FHWA reviewers of a draft of this paper.