Shashi Nambisan Institute for Transportation Iowa State University Murat Ozen
description
Transcript of Shashi Nambisan Institute for Transportation Iowa State University Murat Ozen
MIDDLE EAST TECHNICAL UNIVERSITY
1
Shashi NambisanInstitute for Transportation
Iowa State University
Murat OzenMiddle East Technical UniversityDepartment of Civil Engineering
Mid-Continent Transportation Research Symposium
Ames, IAAugust 15, 2013
An Exploratory Analysis of Large Truck Crashes in the US
MIDDLE EAST TECHNICAL UNIVERSITY
2
Introduction• Truck operations: Key for trade & commerce
– Disproportionate share of fatal crashes• Safety analysis: Macro-level; Fatal crash rates
– Outcomes: Number and type of crashes– Exposure measures: #Trucks, Travel, Type & Extent of Roads
• Analyses of truck safety records: 2000 to 2010• Definitions of trucks before and after 2007 (US DOT)• Fatal truck crash analysis
– Spatial considerations– Temporal considerations
MIDDLE EAST TECHNICAL UNIVERSITY
3
Methodology• Large truck fatal crash
rates per 100 million VKT • Annual rates
– At national level– At regional level for the US DOT’s
10 federal regions
• Descriptive analyses • Statistical tests
MIDDLE EAST TECHNICAL UNIVERSITY
4
Data Sources• US DOT
– Federal Highway Administration (FHWA)– Federal Motor Carrier Safety Administration (FMCSA)
• Large trucks = Gross vehicle weight rating or gross combination weight rating of more than 4.5 tons
• National level fatal crash and truck Vehicle Kilometer of Travel (VKT) data available for the study period
• State level truck VKT data available only 2009, 2010– Average % of national VKT for each state in 2009, 2010– Apply to national VKT for the years 2000 to 2008
National Level Trends• % registered trucks in
national vehicle fleet increased from 3.5% in 2006 to 4.2% in 2007
• Truck VKT increased 36.7% in 2007 as compared to 2006
• Truck fatal crash rate per 100 million vehicle kilometers traveled by large trucks decreased from 1.21 to 0.86
Year FatalCrashes
Million TruckVKT
Fatal Crash Rate per 100 Million VKT
2000 4,573 330,752 1.39
2001 4,451 336,236 1.32
2002 4,224 345,369 1.22
2003 4,335 350,637 1.24
2004 4,478 355,360 1.26
2005 4,551 358,115 1.27
2006 4,350 358,099 1.21
2007 4,204 489,526 0.86
2008 3,754 499,990 0.75
2009 2,987 463,498 0.65
2010 3,261 461,213 0.71
5MIDDLE EAST TECHNICAL UNIVERSITY
MIDDLE EAST TECHNICAL UNIVERSITY
6
Check Data for Normality• Shapiro-Wilk normality test for normality assumption
– Ho: The data are normally distributed– H1: The data are not normally distributed
Period Data Test Statistic df Significance Reject Ho (95%)?
Before 2007
National .877 7 .214 YESUrban .941 7 .649 YESRural .983 7 .974 YES
After 2007
National .981 4 .908 YESUrban .982 4 .911 YESRural .963 4 .796 YES
Shapiro-Wilk Normality Tests Results For National Level Fatal Truck Crash Data
• Need to use non-parametric statistical methods
MIDDLE EAST TECHNICAL UNIVERSITY
7
Non-Parametric Tests• Wilcoxon Rank-Sum test and Mann-Whitney U test
– to test the null hypothesis that two populations are the same• Kruskall-Wallis H test
– to test the null hypothesis that more than two populations are the same
MIDDLE EAST TECHNICAL UNIVERSITY
8
Urban vs. Rural• Are the Urban & Rural fatal truck crash rates the same before and after 2007?
o Before 2007 (i.e. Averagerural = 1.63 > Averageurban = 0.84)o After 2007 (i.e. Averagerural = 0.94 > Averageurban = 0.52)
• Ho: Urban and rural fatal truck crash rates are the same• H1: Urban and rural fatal truck crash rates are NOT the same
Hypothesis Test Type Test Statistics (ts)/ Z-Value
CriticalValue
Reject Ho (95%)
BEFORE 2007W R-S (ts) 28 L=36 YES
M-W U (Z) -3.137 -1.96 YES
AFTER 2007W R-S (ts) 10 L=11 YES
M-W U (Z) -2.309 -1.96 YES
Reject WR-S test Ho if ts < L or ts > UReject M-W U test Ho if absolute value of z > zC
• Rural fatal truck crash rates are higher than urban fatal truck crash rates
MIDDLE EAST TECHNICAL UNIVERSITY
9
Roadway Types• Compare fatal truck crash
rates for interstate highways, other arterials and other highways before and after 2007
• Interstate highways have lower fatal crash rates than other types of roadways.
• Are the differences in fatal crash rates on interstates and other arterials and other arterials and other highways statistically significant?
MIDDLE EAST TECHNICAL UNIVERSITY
10
Roadway Type Comparisons• Ho: Fatal truck crash rates are the same for interstate highways,
other arterials and other highways • H1: Fatal truck crash rates are NOT the same for interstate
highways, other arterials and other highways
Hypothesis Test Type Test Statistics (ts)
CriticalValue
Reject Ho (95%)
BEFORE 2007 K-W H (ts) 17.83 5.99 YES
AFTER 2007 K-W H (ts) 8.35 5.99 YES
Reject K-W H test Ho if H > Hcritical
MIDDLE EAST TECHNICAL UNIVERSITY
11
Regional Trends: Before 2007
• Fatal crash rate in Region 4 is the highest
• Fatal crash rate in Region 10 is the is lowest
• Fatal crash rates in Region 3, Region 8 and Region 10 seem to be lower than the others
Regions(b)
Regions(a)
Fata
l Tru
ck C
rash
Ra
te
Fata
l Tru
ck C
rash
Ra
te
MIDDLE EAST TECHNICAL UNIVERSITY
12
Regional Trends: After 2007
• Fatal crash rate in Region 4 is the highest
• Fatal crash rate in Region 10 is the is lowest
• Variations in the fatal truck crash rates across the regions are not pronounced
Regions(b)
Regions(a)
Fata
l Tru
ck C
rash
Ra
te
Fata
l Tru
ck C
rash
Ra
te
MIDDLE EAST TECHNICAL UNIVERSITY
13
Regional Comparisons: Pairwise• Fatal truck crash rates are not the same for different regions before and after
2007 based on pairwise comparisons.• Ho: Fatal truck crash rates are the same for each region • H1: Fatal truck crash rates are NOT the same for each region
Hypothesis Test Type Test Statistics (ts) Critical
ValueReject Ho
(95%)
BEFORE 2007 K-W H (ts) 55.93 16.92 YES
AFTER 2007 K-W H (ts) 24.73 16.92 YES
Reject K-W H test Ho if H > Hcritical
MIDDLE EAST TECHNICAL UNIVERSITY
14
Regional Ranks: Before 2007
MIDDLE EAST TECHNICAL UNIVERSITY
15
Regional Ranks: After 2007
MIDDLE EAST TECHNICAL UNIVERSITY
16
State Level Analyses• Comparison of state level fatal truck crash rates across the
states before and after 2007• Ho: Fatal truck crash rates are the same across the states• H1: Fatal truck crash rates are NOT the same across the states
Hypothesis Test Type Test Statistics (ts) Critical
ValueReject Ho
(95%)
BEFORE 2007 K-W H (ts)
240.84 66.34 YES
AFTER 2007 K-W H (ts)
126.36 66.34 YES
Reject K-W H test Ho if H > Hcritical
MIDDLE EAST TECHNICAL UNIVERSITY
17
State Level Ranks: Before 2007
MIDDLE EAST TECHNICAL UNIVERSITY
18
State Level Ranks: After 2007
MIDDLE EAST TECHNICAL UNIVERSITY
19
Summary• Illustrated methods to analyze truck safety
– Non-parametric techniques
• Variations in fatal truck crash rates – Before and after 2007– Urban vs rural– Across roadway types– Across federal regions– Across states– Northwest US vs Southeast US
• Limitations– Focused only on fatal truck crash rates– Need to evaluate other factors / metrics
MIDDLE EAST TECHNICAL UNIVERSITY
20
Implications• Spatial and temporal variations in truck safety records warrant
further study• Changes in truck definitions affected estimated crash rates
– Use caution when comparing Pre-2007 and Post -2007 data
• Method can be adapted for safety analyses in other contexts• Support policy making
– Planning– Design– Operations– Law enforcement strategies– Resource allocation– Others
MIDDLE EAST TECHNICAL UNIVERSITY
21
Contact InformationShashi Nambisan, Ph.D., P.E.
Professor of Civil Engineering
Institute for Transportation
Iowa State University
Ames, IA 50011 [email protected]
Murat Ozen, Ph.D. Candidate
Middle East Technical University
Department of Civil Engineering 06800, Ankara - TURKEY