Вы находитесь на странице: 1из 8

JOURNAL OF ADVANCED TRANSPORTATION

J. Adv. Transp. 2010; 44:34–41


Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/atr.104

Severity of urban transit bus crashes in Bangladesh

Upal Barua and Richard Tay*


Department of Civil Engineering, University of Calgary, Calgary, Alberta, Canada

SUMMARY
Unlike in developed countries where buses are a relatively safe mode of transport, there is a significant safety
concern in many developing countries like Bangladesh regarding transit buses. Nevertheless, few studies
have examined the factors contributing to the number or severity of bus crashes. Using the ordered probit
model on bus crash data from 1998 to 2005 in Dhaka, Bangladesh, our study shows that there is a general
increase in the severity of transit bus crashes over this period. Also, crash severity tends to increase when the
collision occurs on weekends, off-peak periods, and two-way streets or involves only one vehicle, a
pedestrian, and other vulnerable road users. On the other hand, the severity of a crash tends to be lower at
locations with some form of police control or road medians, as well as for crashes involving hit object,
parked vehicles, or sideswipes. Copyright # 2010 John Wiley & Sons, Ltd.

KEY WORDS: transit bus; crash severity; ordered probit model; Bangladesh

1. BACKGROUND

The problem of deaths and injuries as a result of road crashes are now acknowledged to be a global
issue with authorities in many countries concerned about the growing number of people killed and
seriously injured on their roads. Around the world, about 1.2 million people are killed every year in
traffic collisions and the problem is expected to get worse, especially in low and middle-income
countries [1]. Likewise, road crashes are a serious and growing concern in Bangladesh where the
situation has been rapidly deteriorating over the years as a direct consequence of rapid growth in
population, motorization, and urbanization. The cost of road traffic injuries was estimated at US$745
million in 2000 or about 1.6% of its gross domestic product [1]. Moreover, unlike many developed
countries where road safety is improving, the number of people killed in traffic collisions in
Bangladesh rose from 1483 in 1993 to 4046 in 2000 [2].
Also, unlike in many developed countries where buses are a relatively safe mode of transport, there
is a significant safety concern in many developing countries like Bangladesh regarding transit buses.
Among all the crashes in Bangladesh from 1998 to 2004, buses were involved in about 15% of them
(MAAP database). Generally, buses are categorized by their purpose and use. In Bangladesh, a transit
bus is used for short-term transportation of people on urban streets, carrying standing and seated
passengers. If only urban areas are considered, transit buses constitute about 17% of all vehicles
involved in crashes. And among all the vehicles involved in fatal crashes in urban areas, about 33% are
transit buses (MAAP database). Therefore, transit buses in urban areas are associated with significant
safety issues in Bangladesh, especially in terms of the severity level involved.

2. LITERATURE REVIEW AND RESEARCH OBJECTIVE

A review of the literature found that very few publicly available studies have been conducted to
examine the crash involvement and severity of transit buses. Most of the analyses on traffic safety were

*Correspondence to: Richard Tay, Department of Civil Engineering, University of Calgary, Calgary, Alberta, Canada.
E-mail: rtay@ucalgary.ca

Copyright # 2010 John Wiley & Sons, Ltd.


URBAN TRANSIT BUS CRASHES IN BANGLADESH 35

concentrated on other modes of transportation rather than transit buses. Among the few studies on bus
crashes, Jacobs and Downing [3] studied the physical condition of buses in Delhi, India, and discovered
rear light faults to be a common problem of buses, which might be a potential cause of crash. In another
study, Hamed et al. (1998) found that the marital status of mini bus drivers, drivers’ interval since
previous crashes, and driving experience had an effect on the occurrence of a crash. Taneerananon and
Somchainuek [4] studied three cases of fatal bus crashes in Thailand and identified three contributing
elements: drivers’ errors, vehicle integrity and defects, and roadside hazards that led to the crashes.
In addition, Zegeer et al. [5] examined the characteristics of commercial bus accidents in five
American states. Using descriptive statistics for 8897 reported bus accidents, they identified several
factors contributing to the occurrence of transit bus crashes including time of day, lighting conditions,
weather and road conditions, bus model year, driver’s age and experience, roadway geometries, and
crash type. Evans and Courtney [6] identified certain factors that were likely to influence occurrence of
bus crashes including time of day, time of year (weekly variations), weather conditions, and driver’s
age and experience, and number of working hours per day. Jovanis et al. [7] analyzed 1800 crashes in
the Chicago metropolitan area that involved mass transit buses and examined the effects of drivers’
age, sex, and experience on the occurrence of a crash. Fruin et al. [8] studied the nature of the injuries
sustained by passengers as they rode, boarded, or exited the bus and recommended changes in bus
design and bus stop location in order to reduce both collision and non-collision injuries. Lastly, White
et al. (1995) found that the deregulation of the bus industry did not have a significant effect on bus
safety in the UK.

3. OBJECTIVE OF STUDY

Our literature review found little research that focus on identifying the factors determining the severity
of transit bus crashes. In addition, no study was found on transit bus safety in Bangladesh even though
bus crashes constituted a very significant share of the crashes, especially fatal crashes. Therefore, this
study aims to identify the factors that contribute to the severity of transit bus crashes to provide
transportation engineers and other road safety professionals with some useful insights to improve
transit bus safety in Bangladesh and other developing countries.

4. DATA AND METHODOLOGY

Traffic crash data in Bangladesh are generally compiled from the First Information Report (FIR)
maintained by the police and subsequently entered into the Micro-computer Accident Analysis
Package (MAAP). Transit bus crashes in Dhaka city have been extracted for this study because about
41% of transit bus crashes in urban areas are concentrated in Dhaka city (MAAP). Of the 6267 crashes
that occurred in Dhaka city from 1998 to 2005, 2662 crashes involved at least one transit bus. Among
these, 39.9% of the cases were classified as fatal, 37.2% were classified as serious injury crashes, 5.2%
were slight injury crashes, and the rest were property damage only crashes (PDO). The severity of a
crash is determined by the person with the most severe injury and a crash is considered as fatal if at least
one person involved in the crash dies, while a serious injury crash is one in which at least one person is
hospitalized. A crash is considered to be a minor injury crash if at least one person involved in the crash
suffers some minor injuries without need of treatment in a hospital. Lastly, a non-injury crash is defined
as a crash associated with no injury but only damage to the vehicles or other properties.
Since data on the severity of crashes are discrete or categorical in nature, researchers have used a
variety of discrete response models. As the severity is also ordinal in nature, many researchers have
chosen to use the ordered probit or logit model to study injury severity because these models yield
estimates that are consistent and efficient [9–17]. However, some researchers are concerned about the
validity of the assumption of a monotonous effect of the independent variables on the dependent
variables, which is implied in an ordered model. These researchers have instead chosen to use
unordered response models including multilevel binomial logit [18,19], multinomial logit [18–21], and
nested logit models [22,23]. In this study, the more widely used ordered probit model will be chosen for
its estimation efficiency since there is no prior evidence to suggest that monotonicity would be a major
issue.

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
36 U. BARUA AND R. TAY

The ordered probit model can be derived from a model in which a latent variable y ranging from
1 to 1 is mapped to an observed ordinal variable y [24]. In this study, y represents the severity of a
transit bus crash with a general specification given by

yi ¼ X i b þ "i (1)

where yi is a latent variable measuring the injury severity of ith collision, Xi is a vector of explanatory
variables, b is a vector of unknown parameters, ei is the Normally distributed error term.
However, the injury level experienced in a crash is often recorded in several ordered categories in
practice. In Bangladesh, the severity of a crash is recorded in four categories: no injury, slight injury,
serious injury, and fatal injury. The latent severity level is thus mapped into an ordinal injury level as
follows:
8
>
> 1 if  1  y < t1 ðNo injuryÞ
<
2 if t1  y < t2 ðSlight injuryÞ
y¼ (2)
>
> 3 if t2  y < t3 ðSerious injuryÞ
: 
4 if t3  y < 1 ðFatal injuryÞ

where t1, t2, t3 are the thresholds separating the four categories of injury severity.
Thus, the estimated probability that road user i sustains an injury of level j ( j ¼ 1, 2, 3, 4) is equal
to the probability that the unobserved injury risk, yi , takes a value within the appropriate ranges shown
above and can be computed as follows:

Prðyi ¼ 1jxi Þ ¼ Fft 1  ðxi bÞg


Prðyi ¼ 2jxi Þ ¼ Fft 2  ðxi bÞg  Fft 1  ðxi bÞg
(3)
Prðyi ¼ 3jxi Þ ¼ Fft 3  ðxi bÞg  Fft 2  ðxi bÞg
Prðyi ¼ 4jxi Þ ¼ 1  Fft3  ðxi bÞg

where F is the cumulative density function for the normal distribution. For all the probabilities to be
positive, it is required that 0 < t1 < t2 < t3. Note that the model is unidentified since a change in the
intercept (b0) in the structural model can always be compensated by a corresponding change in the
thresholds. Hence, either b0 or t1 is usually constrained to be zero and b0 is chosen in this analysis for
convenience. The above model will be estimated using the maximum likelihood procedure in STATA
version 9.
Theoretically, the severity of a crash may depend on many factors, including the type of vehicle units
involved, the speed of the collision, the types of collision, and characteristics of the road users involved.
One way of sorting out these factors is to deliberate upon similar studies where these factors have been
examined. In addition, some local factors that might have a significant influence on crash severity in
Bangladesh are also investigated. However, reliable data on some of these variables are often
unavailable.
This lack of information about road traffic crashes in Bangladesh often compels the researchers to
discard some variables that might be significant in crash analysis. Following these considerations, 13
factors are selected for investigation based on their availability in the crash database. After some
preliminary analyses, eight factors are retained in the final model.
Summary statistics of the variables used are shown in Table I. Note that some of the factors are
categorical in nature and therefore several dichotomous variables are created for each of these factors.
Also, one of the dichotomous variables for each of these factors will not be included in the estimation
model because it is used as the reference case by which the results of the other dichotomous variables
included in the model are compared. For example, when considering the effect of time-of-day, the
reference case used in estimation is peak period. The estimated effects of daytime-off-peak period and
night time are then computed relative to the peak period.
Although most of the factors and variables are self-explanatory, one requires further clarification.
The time of day is divided into the three typical periods. The peak period consists of the morning peak
hours between 9:00 am and 12:00 pm and the afternoon peak hours are from 3:00 pm to 6:00 pm. The
daytime off-peak period comprises the morning off-peak hours between 6:00 am and 9:00 am and the

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
URBAN TRANSIT BUS CRASHES IN BANGLADESH 37

Table I. Explanatory variables used in the models.

Variables Description Mean SD

Time trend Year (1 ¼ 1998 to 8 ¼ 2005) 4.23 2.28


Weekday 1 ¼ weekday; otherwise ¼ 0 0.71 0.45
Time of the day
Off-peak 1 ¼ off-peak; otherwise ¼ 0 0.36 0.48
Night time 1 ¼ night time; otherwise ¼ 0 0.28 0.45
Peak 1 ¼ peak period; otherwise ¼ 0 0.36 0.48
No. of vehicles involved
Single vehicle 1 ¼ single vehicle; otherwise ¼ 0 0.48 0.50
2 Vehicles 1 ¼ 2 vehicles; otherwise ¼ 0 0.50 0.50
3þ Vehicles 1 ¼ 3þ vehicles; otherwise ¼ 0 0.02 0.15
Traffic control and operation
Unsignalized 1 ¼ unsignalized; otherwise ¼ 0 0.57 0.49
Police 1 ¼ police; otherwise 0 0.15 0.44
Signal þ police 1 ¼ signal þ police; otherwise ¼ 0 0.21 0.40
Signal 1 ¼ signalized; otherwise ¼ 0 0.01 0.12
Other controls 1 ¼ other; otherwise 0 0.06 0.24
Types of collision
Head-on 1 ¼ head-on; otherwise ¼ 0 0.05 0.20
Rear-end 1 ¼ rear-end; otherwise ¼ 0 0.32 0.47
Right-angle 1 ¼ right-angle; otherwise 0 0.03 0.18
Side-swipe 1 ¼ sideswipe; otherwise ¼ 0 0.07 0.26
Overturn 1 ¼ overturn; otherwise ¼ 0 0.02 0.13
Parked vehicle 1 ¼ parked vehicle; otherwise ¼ 0 0.03 0.18
Pedestrian 1 ¼ pedestrian; otherwise ¼ 0 0.41 0.49
Fixed object 1 ¼ fixed object; otherwise ¼ 0 0.02 0.14
Others 1 ¼ others; otherwise ¼ 0 0.05 0.12
Traffic flow 1 ¼ two-way; otherwise ¼ 0 0.91 0.29
Median 1 ¼ median is present; otherwise ¼ 0 0.82 0.38

afternoon off-peak hours from 12.00 pm to 3.00 pm. Finally, night time is defined as the hours between
6:00 pm and 6:00 am.

5. RESULTS AND DISCUSSION

The estimation results of the ordered probit model are presented in Table II. In general, the model fitted
the data very well, with a large x2 statistic and a very small p-value for goodness-of-fit. Note that some
of variables are not significant but are retained in the model as long as at least one of the variables for
the same factor is statistically significant. As suggested by some researchers, variables with low
statistical significance may also be retained in the model if they belong to factors that have some
significant effect on injury severity [15]. Although this approach may reduce the efficiency of the
estimates, it is adopted for ease of comparison and interpretation of the estimates. This potential
decrease in efficiency has been adjusted by using a more liberal confidence level of 90% instead of the
traditional 95%.
Time trend over the years was found to have a marginally significant ( p-value ¼ 0.064) relationship
with the severity of transit bus crashes. The positive coefficient of time trend implies that the severity of
transit bus involved crashes has been increasing over time from 1998 to 2005 despite the improved
enforcement, design and traffic control over the years. This result is consistent with a study by Barua
and Tay [11] where fatality risk at urban intersections in Bangladesh is found to be increasing over the
years. These results are consistent with the prediction by the World Health Organization that road
safety in low- and middle-income countries are expected to deteriorate over time [1].
Day of week has been found to have significant effect on the severity of transit bus crashes
( p ¼ 0.020). Crashes on weekends are associated with higher severity risk compared to the transit bus
crashes on weekdays. Usually, the traffic volume in Dhaka on weekends is comparatively lower than
that on weekdays. Thus, the speed of buses on weekends may be higher than on weekdays, which may
be one of the causes of higher severity of transit bus crashes on weekends.

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
38 U. BARUA AND R. TAY

Table II. Parameter estimates of the model.

Variables Coefficient Std. Error z-stat p-value

Year 0.009 0.0106 1.88 0.064


Weekday 0.116 0.0499 2.34 0.020
Time of the day (Reference case: peak period)
Off-peak 0.212 0.0537 3.95 <0.001
Night time 0.111 0.1219 0.90 0.368
No. of vehicles involved (Reference case: 3þ vehicles)
Single vehicle 0.463 0.2108 2.20 0.028
2 Vehicles 0.208 0.1477 1.41 0.159
Traffic control (Reference: unsignalized)
Police 0.284 0.0831 3.41 0.001
Signal .313 0.1936 1.62 0.106
Signal þ police .551 0.0897 6.15 <0.001
Other control 0.262 0.1084 2.42 0.016
Types of collision (Reference: rear-end)
Head-on 0.177 0.1100 1.61 0.108
Right-angle 0.066 0.1289 0.51 0.609
Side-swipe 0.356 0.0917 3.88 <0.001
Overturn 0.261 0.2245 1.16 0.245
Parked vehicle 0.711 0.1354 5.25 <0.001
Pedestrian 0.673 0.1575 4.27 <0.001
Fixed object 1.613 0. 2261 7.13 <0.001
Others 0. 304 0.1713 1.77 0.076
Traffic flow 0.228 0.0838 2.72 0.007
Median 0.161 0.0627 2.57 0.007
Number of observations: 2662; log likelihood ¼ 2640.48; x2 ¼ 1089.68; p-value < 0.0001; pseudo R2 ¼ 0.1710.

Relative to the peak hours, transit bus crashes occurring during off-peak hours are found to be
associated with higher severity risk ( p < 0.001). During off peak hours, because of lower traffic density
on the road, bus drivers and also other motorists may have a greater tendency to speed, thereby
increasing the severity of a crash. On the other hand, night time crashes are found not to have any
significant difference in severity compared to peak hour crashes. Part of the reason for this finding may
be the lower traffic volume during the night, especially pedestrian traffic, thereby reducing the
exposure to risks. This finding is consistent with other studies on Bangladesh where the crashes at
urban intersections [11] and circular intersections [25] during off-peak period and night time are found
to be associated with higher fatality risks compared to daytime crashes. Also, another study by Hamed
et al. [26] shows that commercial mini bus accidents in Jordan have one of the peak periods between 7.00
am and 9.00 am which almost coincides with the morning off-peak period in our study. In addition, Evans
and Courtney [6] also found 9.00 am to be one of the peak periods for bus accidents in Hong Kong.
The number of vehicles involved in a crash has been found to be a significant factor affecting the
severity risks of transit bus crashes. Of the reported transit bus involved crashes, 50% are two-vehicle
crashes, 48% are single-vehicle crashes, and only 2% are three or more vehicle crashes. Single vehicle
crashes include predominantly crashes involving pedestrian, cyclists, and other non-motorized
vehicles such as rickshaws and pushcarts, which is expected to be more severe because these road users
are unprotected and very vulnerable. Other single-vehicle crashes involve hit-fixed-objects and hit-
parked vehicle crashes which are expected to be less serious in Dhaka due to better occupant protection
and relatively low speed compared to many developed countries. Our analysis shows that single-
vehicle crashes are more severe ( p ¼ 0.028) while two-vehicle crashes are similar ( p ¼ 0.159) in crash
severity compared with three or more vehicle crashes. This result is consistent with another study on
commercial mini bus crashes by Hamed et al. [26] where the single-vehicle crashes have been found to
have higher severity risk. It is also consistent with the results of two other studies on crash severity in
Bangladesh [11,25].
Among the different collision types, hit-pedestrian crashes are associated with significantly higher
severity risk ( p < 0.001). This result is expected because pedestrians tend to be more vulnerable due to
the lack of protection and will thus suffer more serious injuries when hit by a motor vehicle. On the

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
URBAN TRANSIT BUS CRASHES IN BANGLADESH 39

Figure 1. Unsafe pedestrian and bus passenger behaviors.

contrary, but also expected, sideswipe collisions, collisions with parked vehicle, and hitting fixed
object are associated with lower severity risks. These findings are consistent with previous studies on
crash severity in Bangladesh [11]. It should be noted that unlike many developed countries, crashes
involving pedestrians is a significant issue in Dhaka due to the unsafe behaviors of many pedestrians
and the poor provision of pedestrians infrastructure [27]. Figure 1 shows some examples of the risky
pedestrian and bus passenger behaviors that are common in Dhaka.
Compared to the unsignalized traffic control scheme, signalized aided with police control, and full
police controlled schemes significantly decrease the severity risk of transit bus crashes. Signalized
control alone does not have any significant effect. These results are consistent with those obtained in
similar studies by Barua and Tay [11] and Tay et al. [25]. Further analysis using only two dichotomous
variables, one each for police and signal control, found that police control was significant whereas
signal control was not. These outcomes imply that the presence of traffic enforcement has a significant
deterrent effect on bus drivers inducing them to drive more carefully and thus decreases the severity
risk in a crash. It also demonstrates that without enforcement, the traffic signal control is not expected
to have an effect on driver behavior. This latter result is also expected because most drivers, including
bus drivers, will ignore the traffic lights if the police are not there to enforce it and sometimes, even
when the police are there because they do not always enforce signal compliance. Figure 2 shows the
traffic at different busy signalized intersections where police are present.

Figure 2. Traffic at signalized intersections.

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
40 U. BARUA AND R. TAY

Traffic flow has also been found to have a significant effect on the severity of transit bus crashes.
Compared to one-way traffic, two-way traffic increases the severity risk of transit bus crashes
( p ¼ 0.007). This result is expected since two-way traffic increases the chances of a head-on collision
which tends to be more severe due to the increase in speed differential as well as hit-pedestrian crashes
which also tend to be more severe due to their vulnerability. This inference is further supported by the
finding that the presence of a median significantly decreases the severity of transit bus crashes. The
latter result is consistent with those obtained by Garder [28].

6. CONCLUSION

Although the transit bus is a primary mode of transport in urban areas, very few studies have been
performed to analyze the safety of transit buses. This study attempts to identify the factors determining
the severity of transit bus crashes in Dhaka, Bangladesh, using the ordered probit model. Our results
show that there is a general increase in the severity of bus transit crashes from 1998 to 2005. The
severity of a transit bus crash increases when the collision occurs on weekends, off-peak periods, or
two-way streets. Severity also increases when the collision involves only one vehicle, a pedestrian, or
other vulnerable road users. On the other hand, the severity of a crash is found to be lower at locations
with some form of police control or road medians, as well as for crashes involving hit object, parked
vehicles, or sideswipes.
The above results suggest that reducing the speed of transit buses during weekends and off-peak
periods, especially where pedestrians are prevalent, may be effective in reducing the severity of
crashes. Also, more driver training and public education aimed at detecting pedestrians and
understanding the vulnerability of pedestrians may reduce the number and severity of not only bus
crashes but crashes involving other vehicles as well. In terms of engineering, the separation of
motorized and non-motorized traffic may be effective in reducing the number of bus crashes and other
types of crashes as well. Also, the provision of traffic medians is likely to be effective in reducing the
severity of bus crashes. Finally, more resources should be devoted to traffic enforcement in general and
police control at intersections in particular, to improve road safety in Bangladesh.

7. LISTS OF SYMBOLS AND ABBREVIATIONS

7.1. Symbols
X independent variables
Y dependent variable
F normal cumulative function
b regression parameters
e error term
t thresholds

7.2. Abbreviations
FIR first information report
MAPP Microcomputer accident analysis package
PDO property damage only crashes
SD standard deviation

ACKNOWLEDGEMENTS

The authors gratefully acknowledge support from the Alberta Motor Association Traffic Safety
Foundation and the Centre for Transportation Engineering and Planning but the views expressed by the
authors do not necessarily reflect those of the organizations. The authors also like to thank the
Bangladesh Traffic Police and Accident Research Centre, Bangladesh for providing the data.

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr
URBAN TRANSIT BUS CRASHES IN BANGLADESH 41

REFERENCES

1. WHO. World Report on Road Traffic Injury, World Health Organization: Geneva, 2004.
2. Hoque M, Alam M, Habib K. Road safety issues and initiatives in Bangladesh: the context of regional significance,
Proceedings of 21st ARRB Transport Research / 11th REAA Conference, Australia, 2003.
3. Jacobs G, Downing A. A study of bus safety in Delhi. TRRL Supplementary Report 758, Transit and Road Research
Laboratory, Crowthorne, U.K, 1982.
4. Taneerananon P, Somchainuek O. Bus crash situation in Thailand: case studies. Journal of the Eastern Asia Society
for Transportation Studies 2005; 6:3617–3628.
5. Zegeer C, Huang H, Stutts J, Rodgman E, Hummer E. Commercial bus accident characteristics and roadway
treatments. Transportation Research Record 1995; 1467:14–22.
6. Evans W, Courtney A. An analysis of accident data for franchised public buses in Hong Kong. Accident Analysis and
Prevention 1985; 17:355–366.
7. Jovanis P, Schofer J, Prevedours P, Tsuno-kowa K. Analysis of bus transit accidents: empirical methodological, and
policy issues. Report UMTA- IL-11-0031-89-01, Urban Mass Transportation Administration U.S. Department of
Transportation, Washington, DC, 1989.
8. Fruin J, Huang H, Zegeer C, Smith N. Recommendations for reducing non collision bus passenger injuries.
Transportation Research Record 1994; 1433:41–48.
9. Rifaat S, Chin H. Accident severity analysis using ordered probit model. Journal of Advanced Transportation 2007;
41(1):91–114.
10. Tay R, Rifaat S. Factors contributing to the severity of crashes at intersections. Journal of Advanced Transportation
2007; 41(3):245–265.
11. Barua U, Tay R. Fatality risks of intersection crashes in Bangladesh. Proceedings of the Conference on Road Safety
on Four Continents, Bangkok, Thailand, 2007.
12. Abdel-Aty M, Keller J. Exploring the overall and specific crash severity levels at signalized intersections. Accident
Analysis and Prevention 2005; 37(3):417–425.
13. Lee C, Abdel-Aty M. Comprehensive analysis of vehicle–pedestrian crashes at intersections in Florida. Accident
Analysis and Prevention 2005; 37(4):775–786.
14. Jianming M, Kockelman K. Anticipating injury & death: controlling for new variables on Southern California
highways. Proceedings of the 83rd Annual Meeting of the Transportation Research Board, Washington, DC, 2004.
15. Kockelman K, Kweon Y. Driver injury severity: an application of ordered probit models. Accident Analysis and
Prevention 2002; 34(3):313–321.
16. Khattak A, Pawlovich M, Souleyrette R, Hallmark S. Factors related to more severe older driver traffic crash injuries.
Journal of Transportation Engineering 2002; 128(3):243–249.
17. Zajac S, Ivan J. Factors influencing injury severity of motor vehicle–crossing pedestrian crashes in rural Connecticut.
Accident Analysis and Prevention 2003; 35(3):369–379.
18. Kim D, Lee Y, Washington S, Choi K. Modeling crash outcome probabilities at rural intersections: Application of
hierarchical binomial logistic models. Accident Analysis and Prevention 2007; 39(1):125–134.
19. Kim J, Kim S, Ulfarsson-Gand-Porrello L. Bicyclist injury severities in bicycle–motor vehicle accidents. Accident
Analysis and Prevention 2007; 39(5):238–251.
20. Savolainen P, Mannering F. Probabilistic models of motorcyclists’ injury severities in single and multi-vehicle
crashes. Accident analysis and prevention 2007; 30(5): 955–963.
21. Neyens D, Boyle L. The effect of distractions on the crash types of teenage drivers. Accident Analysis and Prevention
2007; 39(1):206–212.
22. Abdel-Aty M, Abdelwahab H. Modeling rear-end collisions including the role of driver’s visibility and light truck
vehicles using a nested logit structure. Accident Analysis and Prevention 2004; 36(3):447–456.
23. Lee J, Mannering F. Impact of roadside features on the frequency and severity of run-off roadway accidents: an
empirical analysis. Accident Analysis and Prevention 2002; 34(2):447–456.
24. Long J. Regression Models for Categorical and Limited Dependent Variables. Sage Publications, Inc.: California,
1997.
25. Tay R, Barua U, Rifaat S. Severity of roundabout crashes in Bangladesh, Manuscript under review, 2007.
26. Hamed M, Jaradat A, Easa S. Analysis of commercial mini-bus accidents. Accident Analysis and Prevention 1998;
30(5): 555–567.
27. Barua U, Tay R, Hoque M, Mamun M. Analysis of pedestrian safety on five major arterial roads in Dhaka,
Bangladesh, Canadian Transportation Research Forum, Winnipeg, Canada, 2007.
28. Garder P. Segment characteristics and severity of head-on crashes on two-lane rural highways in Maine. Accident
Analysis and Prevention 2006; 38(4): 652–661.

Copyright # 2010 John Wiley & Sons, Ltd. J. Adv. Transp. 2010; 44:34–41
DOI: 10.1002/atr

Вам также может понравиться