Professional Documents
Culture Documents
1 s2.0 S0001457505002058 Main PDF
1 s2.0 S0001457505002058 Main PDF
Received 3 March 2005; received in revised form 22 September 2005; accepted 5 December 2005
Abstract
This paper examines the temporal transferability of the zonal accident prediction models by using appropriate evaluation measures of predictive
performance to assess whether the relationship between the dependent and independent variables holds reasonably well across time. The two
temporal contexts are the years 1996 and 2001, with updated 1996 models being used to predict 2001 accidents in each traffic zone of the City of
Toronto. The paper examines alternative updating methods for temporal transfer by imagining that only a sample of 2001 data is available. The
sensitivity of the performance of the updated models to the 2001 sample size is explored. The updating procedures examined include the Bayesian
updating approach and the application of calibration factors to the 1996 models. Models calibrated for the 2001 samples were also explored,
but were found to be inadequate. The results show that the models are not transferable in a strict statistical sense. However, relative measures
of transferability indicate that the transferred models yield useful information in the application context. Also, it is concluded that the updated
accident models using the calibration factors produce better results for predicting the number of accidents in the year 2001 than using the Bayesian
approach.
© 2005 Elsevier Ltd. All rights reserved.
Keywords: Temporal transferability; Model calibration; Bayesian approach; Safety planning models
0001-4575/$ – see front matter © 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aap.2005.12.003
580 A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589
appropriate functional model form for calibration of accident dent and independent variables holds reasonably well across
prediction models is a complex task. It gets even more com- time.
plicated for safety planning models because of the paucity of The two temporal contexts are the years 1996 and 2001, with
these. models estimated from 1996 data being used to predict 2001
Data for transportation planning variables are usually accidents. A series of accident prediction models based on 1996
obtained from surveys held periodically (for example, the Trans- database were calibrated in the previous study (Hadayeghi et al.,
portation Tomorrow Survey is conducted once every 5 years in 2003) to estimate the number of accidents in planning zones in
the Greater Toronto Area). These surveys usually cover a com- the City of Toronto as a function of planning variables. It should
prehensive area and are therefore expensive in terms of the time be stressed that the spatial transferability of accident prediction
and cost associated with data collection and analysis. As such, models is outside the scope of this paper. However, some insights
once a model is well specified to capture the decision process in on this issue have been provided in some other papers (Persaud
one context, it is cost effective to transfer such a model to other et al., 2002).
contexts that have limited calibration data as long as the basic The second objective of the paper is to compare alternative
nature of the decision-making process remains the same (Badoe updating methods of temporal transfer and to estimate effects of
and Miller, 1995a). the size of the application context data on model transferability.
There are two aspects to model transferability—temporal and The updating procedures examined are the Bayesian updating
spatial. The first, results from application of a model estimated and updating based on recalibration factors. Models calibrated
at one point of time for prediction at another point of time for the for the 2001 samples were also explored, but were found to be
same spatial environment. This occurs whenever estimated mod- inadequate, so were not tested in the 2001 application context.
els are applied in a forecasting framework. The second aspect In addition, this study investigates the transfer bias caused by
– spatial transferability – involves the application of a model sample size. This issue was not addressed in previous studies
estimated in a specific geographic region for prediction in a dif- which focused mainly on using only one method and one sam-
ferent one for the same time period. Of course, there can be a ple size. This is important since it is believed a large portion of
combination of temporal and spatial transfers. differences in coefficients and predictions is caused by random
Several research efforts have been conducted to address the variation due to selection of data. Also, the goodness of fit mea-
temporal and spatial transferability of transportation planning sures resulting from updated models are compared to the models
models. Most of this research was on trip generation models and estimated using the entire dataset in the application context. A
disaggregate and aggregate mode choice models (Ben-Akiva, key contribution of this paper is the variety in the validation
1981; Koppelman and Wilmot, 1982; Atherton and Ben-Akiva, methods and measures used, tested and compared.
1976; Gunn et al., 1985; Gunn, 2001; Badoe and Miller, 1995b;
Ben-Akiva and Bolduc, 1987; Parody, 1977). 2. Data
However, no similar efforts have been directed at safety plan-
ning models. This is likely in reflection of the reality that few The data used in this study are comprised of accident, socioe-
safety planning models have been calibrated to date. A litera- conomic, demographic and road network characteristics, and
ture search indicated that only few studies have been undertaken traffic flow for the City of Toronto’s 463 traffic zones for 1996
to evaluate and quantify traffic safety impacts in the planning and 481 traffic zones for 2001. The zone definitions for 1996 and
process (Hadayeghi et al., 2003; Lovegrove and Sayed, 2005; 2001 are not identical. For the purpose of this study, the 2001
Ladron de Guevara et al., 2004; NCHRP 8-44, in press). zone definition is adopted for both 1996 and 2001.
Of most relevance to the current effort was some work The data for this study were obtained from three sources.
by Oh et al. (2003) that validated intersection accident pre- The accident data for 1996 and 2001, the first dataset, was pro-
diction models intended for use in the IHSDM from two vided electronically by the Traffic Data Centre (TDC) of the
perspectives—internal validity and external validity. Internal City of Toronto. The accident data are geocoded, which allows
model validity examines the strength of the original models in the grouping of accidents by traffic zone using GIS tools. The
explaining the underlying phenomenon. External validity exam- relevant information for each accident record includes sever-
ines the ability of the models to predict crashes over space and ity (fatal or non-fatal injury and property damage only (PDO))
time. The authors recommended that the validation over time and period of day. The latter allows for separate models to be
could suffer from omitted variables that influence safety, poorly developed for the morning peak period (6:00–9:00 am) to which
measured and surrogate variables, and model functional forms. traffic intensity data pertained.
On the other hand, the authors argue that internal validity and The second dataset, consisting of socioeconomic and demo-
validation over space could suffer from omitted variables, site graphic data, was obtained from the 1996 and 2001 Transporta-
selection and countermeasure selection bias, poorly measured tion Tomorrow Survey (TTS), a 1-day travel survey conducted
and surrogate variables, and misspecification of model func- in the Greater Toronto and surrounding areas (GTA) on approxi-
tional form. mately 5% of its population. The survey is conducted to provide
This paper has two basic objectives. First, it examines the detailed demographic and travel information of GTA house-
temporal transferability of the accident prediction models by hold members to local, regional, provincial road authorities
using appropriate evaluation measures of predictive perfor- and to transit operating agencies. The traffic demand data were
mance to assess whether the relationship between the depen- obtained from the results of a traffic assignment of the morning
A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589 581
peak-period vehicle trips to the road network using the EMME/2 age changes in zonal traffic volume from 1996 to 2001 in the
(1998) software package to estimate vehicle kilometers travelled City of Toronto, based on the 1996 zonal definitions, are pre-
(VKT) in each zone. This approximation is necessary because sented in Fig. 1 . As seen, most of the zones experienced an
cordon counts, the only available source of traffic demand data increase in traffic volume in the 5-year period.
collected consistently across the city, are made at a limited Table 1 also shows that the total number of accidents
number of locations and therefore cannot be used directly and increased by 27% from 1996 to 2001, an average annual increase
sufficiently to provide traffic flow estimates for each zone in the of 5.41%. On the other hand, the number of severe (i.e., fatal plus
city. It is worth mentioning that it is a common practice in build- injury) accidents increased only by a total of 5.64% in the 5-year
ing travel demand and planning models to use traffic assignment period.
model outputs. Fig. 2 presents the zonal percentage change in total number
In addition to the above data, 1996 GIS road files were of accidents for the City of Toronto. As can be seen, most of
obtained from the Integrated Land Use, Transport and Envi- the zones outside of downtown have experienced a higher per-
ronment (ILUTE) group at the University of Toronto, and the centage change in total number of accidents than that for the
corresponding 2001 database, CanMap Street Files, was attained downtown zones.
from University of Toronto Map Library.
4. Model specification
3. Comparison of 1996 and 2001 conditions
A series of macro-level, zonal accident prediction models that
Table 1 presents the magnitude of changes in various char- estimate the number of zonal accidents in the City of Toronto for
acteristics in the City of Toronto between 1996 and 2001. The 1996 was developed in the work described in Hadayeghi et al.
number of households grew by 3.68%, from 908,499 in 1996 (2003). A generalized linear modelling approach was employed,
to 943,252 in 2001. The population of the City of Toronto grew in which negative binomial regression models were developed
by 2.7% in the 5-year period, while the total population of the separately for total accidents and for severe (fatal and non fatal
Greater Toronto Area (GTA) increased by 9.3%. The difference injury) accidents as a function of socioeconomic/demographic,
in percent growth between the GTA and City of Toronto is indica- traffic demand and network data variables. A total of 22 possi-
tive of the reality that more people lived in the suburbs in 2001 ble explanatory variables describing zonal characteristics were
as a result of urban sprawl. During this period, the total employ- explored. The definition of variables was based on (a) previous
ment in the City of Toronto increased by more than 7%. research for similar studies and (b) an iterative process in which
Total vehicle kilometers travelled (VKT) per person various definitions were tried and the ones that gave the best
increased by 15%, from 0.75 in 1996 to 0.88 in 2001. How- results were ultimately selected.
ever, there was little change in total major road kilometers in The following model form was used
the same period, suggesting that traffic conditions should have
worsened from 1996 to 2001. This observation is confirmed by n
the average volume to capacity (V/C) ratios in Table 1, which E(A) = a0 VKTb0 exp bi Xi (1)
increased by almost 30% from 1996 to 2001. The percent aver- i=1
Table 1
Change between 1996 and 2001 data in the City of Toronto
Attribute for independent variables 1996 2001 Difference Percent change (%)
(a) Changes in urban structure, travel behavior and socioeconomic and demographic
Vehicle kilometer travelled (VKT) 1728775 2089257 360482 17.25
Average volume capacity (V/C) 0.22 0.31 0.09 30.18
Average posted speed 42.08 42.48 0.40 0.95
Total population 2305600 2368700 63100 2.66
Number of households 908499 943252 34753 3.68
Number of part-time employees 181224 187890 6666 3.55
Number of full-time employees 875225 943373 68148 7.22
Total number of employees 1056449 1131263 74814 6.61
Total minor road kilometer 4172.66 4406.24 233.58 5.30
Total major road kilometer 859.83 900.73 4.90 0.54
Total number of intersections 15498 16649 1151 6.91
Attribute for dependent variables 1996 2001 Difference Percent change (%)
Fig. 1. Percent change in average traffic volume from 1996 to 2001 in the City of Toronto based on 1996 zonal definitions.
Fig. 2. Percent change in total number of accidents from 1996 to 2001 in the City of Toronto based on 1996 zonal definitions.
A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589 583
Table 2
Results of negative binomial regression for models I and II
Parameter Yearly total accident model (Model I) Yearly severe accident model (Model II)
Intercept (log of α0 ) 4.1767 (0.2957, 199.47) 2.8065 (0.3572, 61.75) 2.6618 (0.3239, 67.55) 1.9593 (0.3736, 27.50)
Log of VKT 0.2656 (0.0347, 58.76) 0.2845 (0.0540, 27.81) 0.2415 (0.039, 38.41) 0.1739 (0.0561, 9.60)
Total major road in kilometer 0.1175 (0.0186, 40.05) 0.1960 (0.0197, 93.98) 0.1225 (0.0202, 36.77) 0.2423 (0.0208, 136.31)
Number of households (× 10−3 ) 0.0864 (0.015, 33.27) 0.0582 (0.0156, 13.89) 0.0858 (0.0164, 27.38) 0.0650 (0.0162, 16.03)
Speed −0.0431 (0.0071, 36.61) −0.0124 (0.0065, 3.64) −0.0312 (0.0078, 16.05) −0.0096 (0.0067, 2.07)
Volume/capacity (V/C) −1.3678 (0.2766, 24.45) −0.6472 (0.3112, 4.33) −1.2232 (0.3042, 16.18) −0.3963 (0.3193, 1.54)
Intersection density 0.0052 (0.0012, 19.87) 0.0024 (0.0012, 3.98) 0.0038 (0.0013, 9.14) 0.0013 (0.0012, 1.03)
Goodness of fit measures
Mean Pearson χ2 /d.f. 1.0313 1.1559 1.0346 1.0110
Pearson R2 37.48% 37.24% 31.90% 43.06%
α 0.1941 0.2355 0.2092 0.2296
Note: Refer to Eq. (1). The dependent variable is the natural logarithm of yearly total accidents. Parameter estimate (S.E., χ2 ).
where E(A) is the predicted accident frequency per zone per year, 5. Measures of model transferability
VKT the zonal vehicle kilometers travelled during peak hours
(6:00–9:00 am), Xi the other explanatory variables, a0 , b0 and bi The transferability of a model relates to the application of
is the model parameters a model estimated in one context based on adequate data to
The above model form is consistent with the non-linear another context with limited data. Model transferability is based
relationship between accidents and traffic intensity (measured on the idea that the estimated model from a previous study in
in this case by VKT) suggested by several other researchers. a different context may provide useful information for estimat-
Separate models are calibrated for total and severe (fatal plus ing parameters appropriate for the application context. “Local”
non-fatal injury) accidents per year and, for each, models were or “estimation” context refers to the context for which the ade-
estimated for the entire day and for the morning peak period quate dataset was used in the model estimation. The “transfer”
(6:00–9:00 am) to which the VKT and other traffic variable or “application” context refers to the context for which predic-
estimates from EMME/2 pertain. The final model parameters tions are required, and to which the model from the local context
and corresponding χ2 and goodness-of-fit criteria are shown in is transferred for application.
Tables 2 and 3. This Table is taken from Hadayeghi et al. (2003) Transportation planning models are not perfectly transferable
which discusses the modelling approach and related issues. between contexts since the models do not capture the behavior
The models developed in this study for both 1996 and 2001 of individuals completely and the data available for the esti-
include major arterial and minor road data in each zone, but mation of the unknown parameters are not error free due to a
exclude freeway data since the presence of a freeway in a zone variety of sources such as sampling errors (Lerman, 1981). As
was found after exploratory analysis to be generally unrelated Lerman (1981) and Ben-Akiva (1981) discuss, it is appropriate
to characteristics of the zone. to describe the measure of transfer effectiveness on a continuous
Table 3
Results of negative binomial regression for models III and IV
Parameter Yearly total accident model for Yearly severe accident model for
morning period (Model III) morning period (Model IV)
1996 Model 2001 Model 1996 Model 2001 Model
Intercept (log of α0 ) 1.7851 (0.2909, 37.66) 0.4328 (0.3127, 1.92) 0.2079 (0.0403, 2.7) 1.0527 (0.4562, 5.32)
Log of VKT 0.2225 (0.0430, 26.73) 0.3268 (0.0574, 32.47) 0.2171 (0.0606, 12.85) 0.3163 (0.0803, 15.53)
Total major road in kilometer 0.1171 (0.0201, 34.07) 0.2127 (0.0202, 111.39) 0.1316 (0.0264, 24.8) 0.2528 (0.0267, 89.51)
Number of households (× 10−3 ) – – 0.0892 (0.0220, 16.43) 0.0701 (0.0216, 10.51)
Number of employements (× 10−3 ) 0.0692 (0.0162, 18.18) 0.0533 (0.0155, 11.80) – –
Speed −0.0324 (0.0073, 19.59) −0.0139 (0.0063, 4.91) −0.0213 (0.0102, 4.34) −0.0141 (0.0087, 2.61)
Volume/capacity (V/C) −1.1156 (0.3296, 11.46) −0.7236 (0.3234, 5.01) −0.9074 (0.4500, 4.07) −0.8137 (0.4412, 3.40)
Intersection density – – – –
Total minor road in kilometer 0.0108 (0.0051, 4.52) 0.0006 (0.0049, 0.02) – –
Goodness of fit measures
Mean Pearson χ2 /d.f. 1.0805 1.0802 1.0791 0.9659
Pearson R2 36.56% 48.46% 31.57% 42.17%
α 0.1586 0.1959 0.1961 0.2290
Note: Refer to Eq. (1). The dependent variable is the natural logarithm of yearly total accidents. Parameter estimate (S.E., χ2 ).
584 A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589
LR = −2[LLi∪j (θ̂i∪j ) − LLi (θ̂i ) − LLj (θ̂j )] (2) 6.1. Updating using a calibration factor
The above equation is χ2 distributed with degrees of freedom The updating procedure was recommended by Harwood et
equal to the number of model parameters. In the event of rejec- al., 2000 and a recent paper (Oh et al., 2003), for application
tion of the null hypothesis, an asymptotic t-test can be performed in the Interactive Highway Safety Design Model (IHSDM) for
to test which parameter(s) were responsible for the rejection. For transferring accident prediction models calibrated for one geo-
this, the t-statistic is given as graphic region to apply to another region. In this procedure,
Harwood et al. (2000) in essence obtained a calibration factor
t = (θ̂ki − θ̂kj )/ (var(θ̂ki ) + var(θ̂kj )) (3) as the total number of accidents observed in a sample from one
jurisdiction divided by the sum of the predicted number of acci-
where θ̂ik is the estimate of parameter k in context I and var(θ̂ik ) dents in the sample using the model from another jurisdiction.
is the variance of the estimate of parameter k in context i. The calibration factor was calculated as follows:
n
n
5.2. Transfer index (TI) calibration factor = Yi / Ŷi (5)
i=1 i=1
Transfer index (TI) is a relative measure that indicates how
well a transferred model performs in predicting the application where Yi is the number of observed accidents in zone i in 2001 for
dataset relative to a locally estimated model in the application a given sample size and Ŷi is the predicted number of accidents
context and of similar specification. For this particular research, in zone i by applying values of the 2001 independent variables
this index compares performance of the transferred and appli- (for the same sample size of 2001) to the 1996 model.
cation context models relative to the performance of a 2001 Then, the calibration factors estimated for each sample size
“constant” model that has no independent variables. The index were simply multiplied by the base models (1996 models) in
has an upper bound, 1.0, which is attained when the transferred order to “transfer” the 1996 models for application to 2001 con-
model performs as well as an estimated model on the applica- ditions.
tion data. It has no lower bound. Negative values imply that the
transferred model is worse than the local constant model. It is 6.2. Bayesian updating
expressed as
Bayesian updating, which was introduced by Atherton and
TI = [LLj (θ̂i ) − LLj (ĉ)]/[LLj (θ̂j ) − LLj (ĉ)] (4) Ben-Akiva (1976), is another approach that is adopted for
updating the 1996 models using the 2001 sample data. This
where LLj (θ̂i ) is the log-likelihood value in context j using methodology combines sample information with prior infor-
model parameters estimated from year i, LLj (θ̂j ) the log- mation in order to achieve more accurate updated information.
likelihood value given by application context model j and LLj (ĉ) Atherton and Ben-Akiva (1976) assumed both prior and pos-
is the log-likelihood value given by constant model estimated in terior (updated) distributions of the parameters to be normally
application context j. distributed. The updated parameters are expressed as
A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589 585
Table 4
Results from transferring 1996 Models to 2001 dataset
Transfer measures Model I Model II Model III Model IV
roadway parameters such as major road kilometers and posted the total accident models (Model III) and severe accident models
speed did not remain stable over time. The value of transfer (Model IV) for the morning peak period. For both Models III
index, which describes the degree to which the log likelihood and IV, the signs of all variables and constant terms are the same
of the transferred model exceeds the reference model relative to for the 1996 and 2001 models. The coefficients in the 2001
improvement provided by a model developed in the application models are largely different from those of the 1996 models, in
context, is found to be −0.966. The negative sign implies that particular, those with low levels of significance such as total
the transferred model is worse than the local reference model. minor road kilometers in the 2001 model. For both Models III
and IV, the coefficient of the total major road kilometers variable
7.1.2. Yearly model for severe accidents for the 2001 models is three times as large as that for the 1996
The model parameters and corresponding χ2 values for the models.
yearly model of severe accidents (Model II) are shown in Table 2. The goodness of fit statistics show mean Pearson χ2 values
All the coefficients are close in magnitude and share the same close to one, again indicating that the models fit the data reason-
sign in the both the 1996 and 2001 models. The intersection ably well. The dispersion parameters (α) for all the models are
density and V/C variables are insignificant at the 5% level in the significantly different from zero, again confirming the appropri-
2001 model but are significant in the 1996 model. ateness of negative binomial regression. The Pearson R2 values
The dispersion parameters (α) of 0.2092, and 0.2226 for the are reasonable for all the models.
1996 model and the 2001 model, respectively, again confirm the The statistical tests of equality of model parameters for the
appropriateness of using negative binomial as opposed to Pois- Models III and IV reject the null hypothesis of equality, indicat-
son regression. The mean Pearson χ2 ratios for the two models ing that some or all model parameters have not remained stable
are within the acceptable range of 0.8–1.2, another indication over time. Both Models III and IV exhibit statistically signifi-
that the choice of the negative binomial error structure appears cant equality at the 5% significance level in all the parameters,
to be appropriate. The Pearson R2 values for the 1996 model except major road kilometers. The values of Transfer index are
and the 2001 model are 31.90% and 43.3%, respectively, which found to be −0.881 and 0.640 for Models III and IV, respec-
indicate that the developed models explain the variation in the tively, suggest that the transferred model provides some portion
dependent variable to a reasonable degree. of the information obtained from the 2001 model, irrespective of
Results of the test of parameter equality in Table 4 indicate model specification. However, the negative value, as explained
that the 1996 and 2001 model parameters are not statistically previously, implies that the transferred model is worse than the
equal. V/C, speed, and major road kilometers are the parameters local reference model.
most responsible for the statistical inequality, as shown in part
(b) of Table 4. The value of transfer index of 0.528, suggests that 7.2. Updating results
the transferred model provides some portion of the information
obtained from the 2001 model, irrespective of model specifica- As expected, none of the models calibrated with limited data
tion. performed reasonably in terms of the magnitude of the coeffi-
cient parameters or their level of significance in the application
7.1.3. Morning peak period accident models context regardless of the sample size. The magnitude of coeffi-
For the morning peak period models, accidents between 6:00 cient parameters for some of the models were the same as the
and 9:00 am were extracted from the dataset by running a query 1996 models but either the parameters were insignificant at the
in Microsoft Access. Table 3 shows the parameter estimates for 5% confidence level or the goodness of fit measures were not
Table 5
Goodness of fit measures for predicting 2001 accidents from 1996, 2001 and updated models
Model I MPB 1.17 −37.80 −10.57 (27.39) −30.05 (39.78) −6.87 (15.49) −28.04 (18.50) 1.31 (17.59) −34.15 (26.86) −2.42 (21.94) −40.64 (23.72)
MAD 59.20 65.17 71.62 (4.86) 61.95 (16.97) 69.73 (2.61) 58.85 (16.14) 71.51 (5.38) 65.19 (18.26) 71.33 (5.61) 65.77 (16.39)
MSE 8067
MSPE 9294 9504 (1690) 8993 (5823) 8956 (294) 9685 (3506) 9066 (688) 8392 (4600) 9205 (673) 8762 (4463)
Model II MPB 2.31 −0.51 −0.66 (8.59) 7.49 (20.33) 0.54 (3.05) −2.69 (5.41) 0.62 (2.12) 0.53 (1.80) −1.17 (4.71) 3.21 (6.00)
MAD 14.89 15.36 17.68 (2.38) 21.45 (13.33) 16.97 (0.56) 15.69 (2.14) 16.78 (0.46) 13.99 (1.87) 16.85 (0.73) 16.58 (2.58)
MSE 449
MSPE 419 545 (123) 808 (902) 482 (12) 445 (127) 476 (4) 513 (90) 494 (25) 441 (180)
Model III MPB 2.28 −6.10 −1.42 (3.56) −6.09 (3.52) −0.53 (1.77) −6.54 (1.78) 1.16 (1.90) −8.24 (2.38) 0.26 (2.51) −5.87 (3.13)
MAD 8.11 8.51 8.47 (0.77) 7.49 (3.21) 8.24 (0.36) 8.74 (1.63) 8.62 (0.63) 10.16 (1.57) 8.48 (0.70) 8.21 (2.10)
MSE 132
MSPE 168 152 (12) 138 (155) 140 (4) 143 (50) 142 (9) 278 (149) 143 (11) 178 (149)
Model IV MPB 0.39 −0.52 0.23 (1.11) −0.76 (1.08) −0.04 (0.85) −0.28 (0.69) 0.10 (0.66) −0.48 (1.06) −0.25 (0.58) −0.29 (0.79)
MAD 2.39 2.34 2.65 (0.16) 2.66 (0.62) 2.54 (0.18) 2.32 (0.75) 2.53 (0.13) 2.50 (0.38) 2.47 (0.10) 2.23 (0.39)
MSPE 10.15
MSE 10.90 12.58 (1.09) 12.89 (7.06) 12.02 (0.65) 11.16 (7.35) 11.72 (0.42) 10.46 (3.98) 11.73 (0.46) 9.22 (3.79)
Note: MPB = mean prediction bias, MAD = mean absolute deviation, MSPE = mean squared prediction error, MSE = mean squared error (S.D.).
587
588 A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589
satisfactory. This was expected since the sample sizes selected This research looked only at temporal transferability of zonal
for this study were too small (less than 8% of the total data) to accident prediction models. A natural next step for further
result in a meaningful conclusion. Therefore, it was concluded research would be the investigation of spatial transferability and
that the updating procedure is the best approach when limited the combination of temporal and spatial transferability. That
data are available and no further investigation was conducted research should aim to identify the conditions that make two
regarding developing models with limited data. study areas “similar enough” to justify transferring a model from
Table 5 presents the relative goodness of fit measures for pre- one to the other.
dicting the number of accidents in 2001 using the 1996 models,
the 2001 models, models that apply a calibration factor to the Acknowledgment
1996 model and the Bayesian approach. For the latter two, vari-
ous sample sizes are used to estimate the “updated” models that The authors acknowledge the Natural Sciences and Engi-
are applied. neering Research Council of Canada (NSERC) for its support
As expected, the measures indicate that 2001 model fits the of some of the fundamental research on which this work is based.
2001 data better than other models. The real value of these results
is in the comparison of the performance of the “updated” models References
among themselves for different sample sizes and with the 1996
model. This comparison shows that in general the recalibrated Atherton, T.J., Ben-Akiva, M.E., 1976. Transferability and updating of dis-
“updated” models perform better than the updated Bayesian aggregation travel demand models. Transport. Res. Rec. 610, 12–18.
models and the1996 models. However, no clear-cut conclusions Badoe, D.A., 1994. An investigation into the long range transferability of
work-trip discrete mode choice models. PhD Dissertation. Department of
can be drawn regarding the performance of updated models in Civil Engineering, University of Toronto, Ont., Canada.
respect to the sample size used for estimating them. Badoe, D.A., Miller, E.J., 1995a. Analysis of the temporal transferability of
disaggregate work trip mode choice models. Transport. Res. Rec. 1493,
1–11.
8. Conclusions and summary Badoe, D.A., Miller, E.J., 1995b. Comparison of alternative methods for
updating disaggregated logit mode choice models. Transport. Res. Rec.
1493, 90–100.
This paper examined the temporal transferability of zonal Ben-Akiva, M., 1981. Issues in transferring and updating travel-behaviour
accident prediction models for the City of Toronto. In so doing, models. In: Stopher, P.R., et al. (Eds.), New Horizons in Travel Behaviour.
model calibration and related transferability issues were also Lexington Books, Lexington, MA, pp. 665–686.
addressed. The coefficient estimates for the 2001 models are Ben-Akiva, M., Bolduc, D., 1987. Approaches to model transferability and
updating: the combined transfer estimator. Transport. Res. Rec. 1139,
consistent with the 1996 models. Most of the coefficient param- 1–7.
eters and constant terms are significant for the 1996 and the 2001 EMME/2, 1998. User’s Guide Manual Release 9. INRO Consultants Inc.
models across all four accident type models (Models I–IV). Gunn, H.F., 2001. Spatial and temporal transferability of relationships
The results of nested likelihood ratio test indicate that the between travel demand, trip cost and travel time. Transport. Res. Part
model parameters from two urban contexts are not statistically E: Logist. Transport. Rev. 37 (2/3), 163–189.
Gunn, H.F., Ben-Akiva, M.E., Bradley, M.A., 1985. Tests of the scaling
equal and have not remained stable over time. Therefore, from approach to transferring disaggregate travel demand models. Transport.
a theoretical perspective, the transfer of macro-level accident Res. Rec. 1037, 21–30.
prediction models is undesirable. However, relative measures of Hadayeghi, A., Shalaby, A.S., Persaud, B., 2003. Macro-level accident pre-
transferability indicate that the transferred models yield useful diction models for evaluating safety of urban transportation systems.
information in the application context. Transfer index values Transport. Res. Rec. 1840, 87–95.
Harwood, D.W., Council, M., Hauer, E., Hughes, W. E., Vogt, A., 2000.
show that transferred models capture at least 52% of the variance Prediction of the expected safety performance of rural two-lane highways.
provided by estimating all the parameters in the 2001 data. The Report FHWA-RD-99-207, FHWA, Department of Transportation.
results of the t-test for all the models indicate that the VKT Koppelman, F.S., Rose, G., 1983. Geographic transfer of travel choice mod-
and socioeconomic and demographic parameters are stable over els: evaluation and procedure. In: Optimization and Discrete Choice in
time. Urban Systems: Proceedings of the International Symposium on New
Directions in Urban Systems Modeling, pp. 272–309.
In the absence of a complete dataset for calibrating the 2001 Koppelman, F.S., Wilmot, C.G., 1982. Transferability analysis of dissagregate
model, it looks like the “updated” models, which are developed choice models. Transport. Res. Rec. 895, 18–24.
by applying calibration factors to the 1996 models, perform Ladron de Guevara, F., Washington, S.P., Oh, J., 2004. Forecasting crashes at
reasonably and better than the 1996 models and the Bayesian the planning level: simultaneous negative binomial crash model applied
updated models regardless of the sample size used for esti- in Tucson, Arizona. Transport. Res. Rec. 1897, 191–199.
Lerman, S.R., 1981. A comment on interspatial, intraspatial, and temporal
mating these factors. The Bayesian approach has not yielded transferability. In: Stopher, P.R., et al. (Eds.), New Horizons in Travel
good results in general. The possible problem for this method Behaviour. Lexington Books, Lexington, MA, pp. 628–632.
is that in this method the coefficients in model transferring are Lovegrove, G.R., Sayed, T., 2005. Macro-level collision prediction models
heavily affected by the estimation context data since the num- for evaluating neighbourhood traffic safety. In: Proceedings of the Sixth
ber of observations is larger in this context. Also, this method Transportation Specialty Conference, Canadian Society of Civil Engineer-
ing, Toronto, Canada.
assumes that the prior and posterior (updated) distributions of Oh, J., Lyon, C., Washington, S., Persaud, B., Bared, J., 2003. Validation
the parameters are normally distributed, which may not be a of FHWA crash models for rural intersection lessons learned. Transport.
good assumption. Res. Rec. 1840, 41–49.
A. Hadayeghi et al. / Accident Analysis and Prevention 38 (2006) 579–589 589
Parody, T.E., 1977. An analysis of disaggregate mode choice models in pre- Persaud, B., Lord, D., Palmisano, J., 2002. Calibration and transferability of
diction. Transport. Res. Rec. 649, 23–31. accident prediction models for urban intersections. Transport. Res. Rec.
National Cooperative Highway Research Program (NCHRP), in press. Incor- 1784, 57–64.
porating Safety into Long-range Transportation Planning, Project 8-44. SAS Institute Inc., 1999. SAS/STAT® Users Guide, Version 8. SAS Institute
National Cooperative Highway Research Program. Inc., Cary, NC.