2015 Transportation Research Board (TRB) Data Contest
The main objective of the 2015 Transportation Research Board Data Contest is to provide a platform for
researchers, practitioners and students to learn about cutting edge statstictical methods employed in
the transportation field. The TRB Contest allows for a fair comparison of various statistical methods on
a common dataset.
Submissions are requested from all communities – researchers, practitioners and students. The authors
can use any statistical methods and/or statistical program for the analysis exercise.
The 2015 TRB data contest is sponsored by the Artificial Intelligence and Advanced Computing Committee
(ABJ70).
The contest is also co-sponsored by Statistics (ABJ80), Urban
Transportation Data and Information Systems Committee (ABJ30),
Committee on National Transportation Data Requirements and Programs (ABJ10), Transportation Demand Forecasting (ADB40) and
Traveler Behavior and Values (ADB10)
For the 2015 TRB data contest we have identified research datasets on (1) Transportation Safety and (2)
Travel Behavior. The authors can choose to participate in one of the two submissions.
The submission and evaluation criterion are identical for both datasets and are provided later.
The details of each dataset are provided below separately for the two contests. It is important that
authors follow the updates to their chosen dataset carefully.
Transportation Safety
Data Description
Data Excel Sheet summarizes 10 years of crash data at midblock segment of arterial roads (urban
collectors, urban minor arterial, urban principal arterial-other non-connecting link, and urban
principal arterial-other connecting link) in 4 cities of Nebraska.
The data set contains segment details measured using Google Earth, such as, lane width, speed limit,
presence of shoulders, etc., and the yearly crash frequency reported for different categories, such as,
crash severity, driver age etc.
The datafile contains 19600 rows of data and 135 columns (columns A to EE). The description of the
variables is provided in the Excel sheet as well.
Competition Objective
Develop an exploratory, analytical or statistical model using the data available in
“TRB2015-DataSet.xlsx” to assess the impact of Narrow Lane Width on safety of the arterial roads. You
can use any statistical/analytical software program.
Data and Discussion Files
For answers to Queries and Comments from authors please see the file below
Questions or Comments?
Please check the website regularly to see if there are any updates or comments/clarification on the
dataset. Please direct all questions related to the Transportation Safety Dataset to Anuj Sharma
(anujpals@gmail.com) or Linda Ng Boyle (linda@uw.edu).
We will post responses to all questions on the website (we will NOT provide individual responses).
Please note that we WILL NOT answer any questions on what the best model is, what is the right goodness
of fit test, what are the model assumptions, how to compute X, Y or Z, etc.
Travel Behavior
Data Description
The data provided corresponds to the New York-Northern New Jersey-Long Island, NY-NJ-PA region from the
2009 National Household Travel Survey (NHTS) Data. The data and associated documentation has been
downloaded from http://nhts.ornl.gov on May 8th 2014 and has been appropriately processed to correspond
to the New York-Northern New Jersey-Long Island region.
Competition Objective
The primary objective of the NHTS data is to enhance our understanding of travel behavior. The data
contest based on the data provided is focused on developing quantitative models that allow us to better
understand and thereby forecast future transportation decisions. The specific objective of the contest
is to examine short term vehicle use decisions. The authors can use any statistical method and/or
statistical program to analyze the data provided. The authors are encouraged to use at least 10% of the
sample for a validation exercise.
Data and Discussion Files
For answers to Queries and Comments from authors please see the file below
Questions or Comments?
Please check the website regularly to see if there are any updates or comments/clarification on the
dataset. Please direct all questions related to the Travel Behavior Dataset to Naveen Eluru
(naveen.eluru@mcgill.ca) or Linda Ng Boyle (linda@uw.edu).
We will post responses to all questions on the website (we will NOT provide individual responses).
Please note that we WILL NOT answer any questions on what the best model is, what is the right goodness
of fit test, what are the model assumptions, how to compute X, Y or Z, etc.
Submission Format
Your results along with a SHORT write up should be NO MORE than 6 pages total (this includes figures,
tables, and references). The file should contain the following
a. Your name(s) and affiliation(s)
b. Your problem formulation
c. Your model and justification for your approach
d. Model adequacy check: How do you know you have a good fitting model?
e. Your solution
f. Your assumptions
g. The software used (and corresponding program or functions/call out procedure)
h. The level that you reduce/aggregate the data for analysis (if any)
i. The limitations in the dataset (including what variables you wish you had).
j. A critical review of your solution process in terms of strengths and weaknesses
It needs to include your Last Name and First Name in the filename and the research dataset (e.g,
Eluru_Naveen_Safety.pdf for safety dataset and Eluru_Naveen_Behavior for behavior dataset).
Entries will be judged based on completeness, applicability, and presentation quality. Entries must be
submitted via email before 11:59 PM, November 30th.
Winners will be notified by end of December or early January.
NOTE: We will NOT review those documents that do not adhere to the submission requirements.
Submission Deadline
Deadline for submission of results and short paper: November 30th, 2014 (11:59 PM, Central Time).
The entries can be emailed to Linda Ng Boyle (linda@uw.edu), Anuj Sharma (anujpals@gmail.com) and Naveen
Eluru (naveen.eluru@mcgill.ca).
Judging Criterion
Entries will be judged based on completeness, applicability, and presentation quality. The TRB contest
submissions will be reviewed by the Contest Board and selected authors will be invited to make
presentation as part of the TRB Data Contest Workshop.
The selected author submissions will also receive awards from the sponsoring committees. The Board
membership includes Linda Ng Boyle (linda@uw.edu), Konstantina Gkritza (nadia@purdue.edu), Jan-Mou Li
(lij3@ornl.gov), Maya Abou Zeid (ma202@aub.edu.lb), Anuj Sharma (anujpals@gmail.com) and Naveen Eluru
(naveen.eluru@mcgill.ca).
Results
The quality of the submissions was very good with a wide range of statistical methods and software
employed. Based on the submissions, the following four awards are given based on their high rankings for
data exploration, model quality, scientific rigor, paper quality, editorial consistency and overall
presentation of results.
Best Paper Safety Area
Title: Examining the Safety Impacts of Narrow Lane Widths on Urban/Suburban Arterials
Authors: Timothy Barrette, Georges Bou-Saab, Amrita Goswamy, Raha Hamzeie, Emira Rista, Brendan Russo, Bo
Wang, Peter Savolainen, Iowa State University
Honorable Mention Safety Area
Title: Quantifying Impacts of Narrow Lane Widths on Arterials Roads
Authors: Zhe Sun, Zhe Yu, and Hernando Ombao, University of California Irvine
Best Paper Travel Behavior Area
Title: Household-Level Vehicle Resources Allocation Model for Analyzing Short Term Vehicle Use Decisions
Author: Rajesh Paleti, Old Dominion University
Best Student Paper
Title: The Relationship between Road Safety and Lane Width: Analysis using Propensity Scores
Author: Yaoyuan Vincent, University of Michigan