category: literaturenote citekey: gundogduapplyinglinearanalysismethods2010 title: "Applying linear analysis methods to GIS-supported procedures for preventing traffic accidents: Case study of Konya" authors: "Gundogdu, Ismail Bulent" year: 2010 date: 2010-00-00 2010 doi: 10.1016/j.ssci.2010.02.016 publication: Safety Science url: "https://doi.org/10.1016/j.ssci.2010.02.016" zotero_key: NN23P7ZA zotero_storage: 5D9PX62X collections: transpordi_planeerimine folder: 001_artiklid/Road Safety Indicators firstAuthor: "Gundogdu, Ismail Bulent"

status: converted

Contents lists available at ScienceDirect

Safety Science

journal homepage: www.elsevier.com/locate/ssci

Applying linear analysis methods to GIS-supported procedures for preventing traffic accidents: Case study of Konya

Ismail Bulent Gundogdu *

Department of Geomatic Engineering, Selcuk University, Konya, Turkey

article info

Article history: Received 26 August 2009 Received in revised form 10 February 2010 Accepted 20 February 2010

Keywords: Linear analysis GIS Traffic accident Highest Probable Hot Spot Hot Pieces

abstract

This study develops methods to obtain maps to determine traffic Hot Spots in Konya, Turkey, by applying linear analysis supported by Geographical Information Systems (GIS).

Hot Spot analysis is known method but the study differs from former researches at the point of determining of risky zones, classification and illustration of them on the maps with the different accident parameters. The aim is not to contribute another Hot Spot analysis using a number of statistical methods, but to determine the Highest Potential Hot Spots (HPHS), which are inter-sectional clusters, and to use different parameters, such as number of accidents, number of fatalities and injured, and number of accidents with only financial loss.

In addition, apart from classical illustrational techniques, Hot Pieces (HPCS) on roads divided into 1 km segments are shown by their grading according to their numerical values. Hence, thematic illustration distinguishes them from others.

Another aspect of this study is that, besides investigation of Hot Spots by means of data of accidents of previous years', Probable Hot Spots (PRHS) were illustrated and highly potential Hot Spots were determined. These latter are candidates for Hot Spots in the near future. Thus, premature accidents can be anticipated easily.

The main intention of this study is to emphasize the importance of using criteria, other than total accident number, to illustrate intersection Hot Spots and to constitute a model of accident severity and variety. It is anticipated that the results obtained from highway accidents data will guide improvement of the route segments.

1. Introduction

Traffic accidents have been recognized as one of the major causes of human and economic losses, both in developed and developing countries. This problem is of great concern in developing countries, because of its seriousness and the limited resources available to develop feasible countermeasures for reducing this ever-growing challenge (Berhanu, 2004).

A report from the World Health Organization and the World Bank (2004) on road traffic accidents and injuries estimated that 1.2 million people are killed in road crashes each year and as many as 50 million are injured worldwide (Rathinam et al., 2007; Rutter and Quine, 1996; Hilakivi et al., 1989). According to estimations of the World Bank, traffic accidents will be the third most frequent cause of deaths in 2020.

Statistics from the Turkish Statistical Institute (TUIK) reveal that, in 2006, there were 665, 618 traffic accidents in Turkey, 2629 of which resulted in mortality, 77,644 resulted in injuries and 585,345 in economical damage. The damage was estimated as \$870 million.

According to a questionnaire from Arslan (2007) among 800 people whose relatives lost their lives in traffic accidents in European countries revealed that 37% of those had a tendency to commit suicide and 64% of those had depression in the subsequent 3 years. Another questionnaire applied in Turkey revealed that 50% of 240 people whose relatives lost their lives in traffic accidents suffered from insomnia and 39.2% suffered from hysteria, in addition to depression.

Several precautionary procedures have been conducted to prevent or minimise the destructive problem of traffic accidents. Plotting, which needs to be conducted by analysing causes of accidents through a large amount of data, is one of the major precautionary measures that might prevent accidents.

According to a report from the Turkish Ministry of Public Works and Settlement (2007), there was, between 2003 and 2005, a 64% decrease in the number of accidents, a 94% decrease in road fatalities and a 65% decrease in the number of people injured on rehabilitated segments of roads.

^* Tel.: +90 555 7214769; fax: +90 332 2410635. E-mail address: bgundogdu@selcuk.edu.tr

The most frequent method in defining a Hot Spot is how they are determined. A Hot Spot can be defined as a point where there is a particularly frequent recurrence of an event. These spots are determined statistically and have various visual methods and usage fields that can be described as follows (Eck and Weisburd, 1995). Areas of concentrated crime are often referred to as Hot Spots. Researchers and police use the term in many different ways. Some refer to Hot Spot addresses (Eck and Weisburd, 1995; Sher man et al., 1989), others refer to Hot Spot blocks (Weisburd and Green, 1994; Taylor et al., 1984), and others examine clusters of blocks (Block and Block, 1995).

The number of accidents on a given road section during a certain period of time is probabilistic in nature and is a non-negative integer. Despite the fact that accidents are random and unpredictable at micro-level, statistical models can predict reliable estimates of expected accidents by relating aggregates of accidents to various explanatory measures of flow, site characteristics, and road geometry at macro-level. Numerous empirical relationships between vehicle accidents and these explanatory variables have been established in several previous studies (e.g. Maher and Summersgill, 1996; Miaou, 1994; Miaou and Lum, 1993). Such accident predicting models are useful in identifying the most critical variables of safety, assessing design and management alternatives, and improving safety standards for new roads.

The significantly different areas detected by the applied statistics (usually termed LISA: Local Indicators of Spatial Autocorrelation) are, traditionally, termed Hot Spots, hotpoints or, occasionally, hotbeds (Ratcliffe and McCullagh, 1999). The statistics used to define the hotbed coverage may result in circular, elliptical or amoeboid shapes depending on the method or application used. There is a wide choice of approaches available (Mccullagh, 2006). The numerical analysis of spatial point distributions has interested academics for many years. A very early work that resulted is the famous Moran's I statistic by Moran (1948). An early important collection of work on general spatial analysis is that edited by Berry and Marble (1968). The development of spatial point analysis theory is illustrated by later works by Cliff and Ord (1973, 1981), Ord and Getis (1995), Anselin (1995) and Gatrell et al. (1996). The overview by Getis and Ord (1996) has led to large number of recent studies, such as Cho (2003), using the techniques that have been developed over the last 50 years (Mccullagh, 2006).

The Linear Referencing Model (LRM) was used to determine Hot Spots in this study.

An LRM can be defined as a mechanism for finding and stating the location of an unknown point along a network by referring it to a known point (Vonderohe and Chou, 1997). More specifically, an LRM is a process for determining a previously unknown location based on a defined path along the underlying transportation network; a distance along that path measured from a known datum location; and, optionally, an offset from the path. There are several types of LRMs, which differ based on the parts of the network used for referencing and the ways in which measures and offsets are calculated (Nyerges, 1990).

The following tasks must be undertaken in a study which has the same aim:

- Determining the locations where an increased number of accidents occur.
- Detailed functional analysis of a location where an accident has occurred for the purpose of determining the factors causing accidents.
- Developing methods used for determining elements, such as traffic notices, and lights controlling traffic flow that cause danger.
- Inferring the effect of physical features of a road such as curve, pavement material, and turning directions on accident occurrence.

Fig. 1. General map of roads in Konya Province and location of the related highways.

Handling traffic problems of the province in a comprehensive context by assessing all highways.

It is not adequate processing just the total number of accidents in carrying out these tasks. All criteria must be evaluated separately. For this purpose, there must be an acceptable and detailed accident report for the long term. It may be possible to explain the effects of different accident information details by comparing four base criteria. These main criteria, detailed below, are the total number of accidents, fatalities, injured and with only financial loss. In particular, result base maps related to base criteria have been produced for this purpose. In addition, the question of how many and which segments are problematic? Will be explained. That the problems are either changing according to period of year or structure of road will be determined.

2. Study area and accident data

2.1. Study area

Highway transport is very important in Turkey. Konya is the biggest city of the country, and has 2957 km of highway, out of a total 61,939 km of roads.

According to statistics in 2006, the city is eighth in the rank of total accident rates, third in the rank of mortality rates, fifth in the rank of injuries. The region has been one of the most important transport connection points for some time. The city is in the centre of the country and connects crowded cities and tourist regions. A large amount of agricultural crops are carried from east to west by trucks with the help of seven routes of the city.

Having a smooth topography and straight routes enables high– speed traffic. Thus, in addition accidents are inevitable, and the severity of accidents and losses may be at a very disturbing level. Fig. 1 shows digitized routes on the map.

2.2. Traffic accident data

The most consuming phase of the study is detailed registration of accident data. In this study, 3856 accidents data for the last 10 years, obtained from the Konya Police Department. All detailed data have been prepared by using MS Access to evaluate by Arc GIS 9.2 program. Seven highways totally 724 km have been digitized with the scale of 1/25,000 maps and connected with coordinates of accident points and other details by the program.

Because of the changing transport density in certain seasons, it will be possible to evaluate old data with new data or a different period of year and different purposes for the future.

3. Methodology

It can be easily understood, through statistical tests, whether or not the inputs are meaningful in determining Hot Spots. The Poisson distribution technique is commonly used for this purpose. Recent studies (e.g. Meel and Safr, 2008) have looked at on application of the Poisson distribution technique to traffic accidents, analysing high-crash locations (Kristen and Mark, 2002) black spots and Poisson distribution (Ding and Pei, 2007) explaining dangerous locations with four categories (Hu et al., 2007), crash counts and prediction modelling (Ma et al., 2008; Caliendo et al., 2007; Mitra and Washington, 2007; Ng et al., 1995). Poisson assumptions for the crash densities (Li et al., 2008) and time series (Quddus, 2008) can be arranged.

The main features of Poisson distribution are that it was developed for specific circumstances of randomly occurring events over a time and for repetitive experiments of events with low potential for occurrence. When formation of Hot Spots is analysed, it is seen that independently occurring individual accidents form Hot Spots by assembling in clusters. This is why Poisson distribution is the most fitting statistical distribution for determining Hot Spots by handling randomly occurring traffic accidents.

In order to use Poisson distribution for the purpose of determining Hot Spots along a road, the road must be divided into certain segments. The area where an accident happened is recorded officially to the nearest kilometre. It is, therefore, sensible to divide the road into 1 km segments. After accident probability for each individual segment is calculated, 1 km segments exceeding the threshold value are characterized as Hot Spots. In this case, the number of accidents is calculated for each part using the formula below (n: number of accidents; k: number of segments).

$$\lambda = \frac{n}{k}$$

Accident potential for each segment is estimated using the formula

$$f(k;\lambda) = \frac{\lambda^k e^{-\lambda}}{k!}$$

With a probability rate of 0.05, where e is the base of the natural logarithm (e = 2.71828...), k is the number of occurrences of an event – the probability of which is given by the function, k! is the factorial of k, and k is a positive real number, equal to the expected number of occurrences that occur during the given interval.

In order to determine which pieces, whose accident potential has previously been estimated with Poisson distribution, have the character of Hot Spots, the Getis–Ord Gi analysis must be applied. The statistics Gi and Gi, introduced by Getis and Ord (1996) for the study of local patterns in spatial data, were extended and re-written (below) in 1995 to redefine Gi as a standard variate and to allow for non-binary classifications of the distance d (Ord and Getis, 1995). In this analysis, the Gi(d) statistic is used as follows (Khan et al., 2005):

$$G{i}(d) = \frac{\sum{j=1}^{N} w{ij}(d)x{j} - \bar{x}{i}\sum{j=1}^{N} w{ij}(d)}{S(i)\sqrt{\frac{\left[\binom{N-1}{j-1}w{ij}^{N}(d) - \left(\sum{j=1}^{N} w{ij}(d)\right)^{2}\right]}{\frac{N-2}{N-2}}}, j \neq i$$

where

$$\bar{x}i = \frac{\sum{j=1, i \neq j}^{N} Xj}{N-1} \quad \text{and} \quad S(i) = \sqrt{\frac{\sum{j=1, i \neq j}^{N} X_j^2}{N-1} - \left(\bar{x}_i\right)^2}$$

$$G{i}^{*}(d) = \frac{\sum{j=1}^{N} w{ij}(d)x{j} - \bar{x}\sum{j=1}^{N} w{ij}(d)}{S\sqrt{\frac{\left[\sum{j=1}^{N} w{ij}^{2}(d) - \left(\sum{j=1}^{N} w{ij}(d)\right)^{2}\right]}{N-1}}}$$

where

$$\bar{x} = \frac{\sum_{j=1}^{N} xj}{N}$$ and $S = \sqrt{\frac{\sum{j=1}^{N} x_j^2}{N} - (\bar{x})^2}$

where Gi is the Getis–Ord Gi z-score value excluding the value at segment i, G ⁱ is Getis–Ord G ⁱ z-score value including the value at segment i, xi is relative crash rate of segment i, xj is relative crash rate of neighboring locations to segment i, d is fixed band radius around segment i, wij is spatial weight matrix for all segment j within distance d, and N is the number of weighted points.

Getis and Ord's Gi analysis calculates the 'z' value for each segment. If the number of accidents at a certain point significantly differs from the z values calculated from the median, z values must be

Fig. 2. Hot Spots are shown according to: (a) number of accidents, (b) rate of mortality, (c) number of people injured, (d) quantity of damage, and (e) the inter-sectional cluster of all showing the highest risk points.

over the threshold value of 1.645. Thus, Hot Spots can be easily determined by an enquiry based on z values.

3.1. Analysis of Hot Spots with unusual criteria and the Highest Potential Hot Spots

By means of this method, at which point and according to which criterion rehabilitation of roads must be performed can be predetermined, Fig. 2 shows Hot Spots according to different criteria. Hot Spots have been determined by the number of accidents so far. The upper limit of analysis (1.645) has been processed in the accident number column. It must be mentioned that other criteria are as important as accident numbers. These are the number of fatalities, number of injured and amount of damage. In this study, Hot Spots were determined on the basis of not only the number of accidents, but also the number of people killed/injured and the amount of damage. This is so that a more realistic analysis can be suggested. Fig. 2a shows the classical method used with accident numbers, but the numbers and sizes of Hot Spots vary with other criteria, as seen in Fig. 2b–d. It must not be forgotten that different maps enable different analyses to take into consideration different criteria. Another important point relates to the intersection segments of the four criteria, as seen in Fig. 2e. The results of this study indicate that, between 10–14 km and 55–65 km of route 3, 25–29 km of route 4 and 37–45 km of route 6 are intersection segments and they must be considered as priority areas.

3.2. Illustration of hazardous segments as Hot Pieces

Maps are indispensable products of numerical data to illustrate or analyse various subjects. According to the type of map reader, or subject, different maps are produced for different purposes. Use of correct illustration techniques allows correct analysis. Occasionally, spot representation may suit any subject, but linear or area may not be sufficient. So, in this study, various maps have been produced by different illustrating types for future investigations.

Hot Spots have been shown in speckles so far. When involved in traffic accidents, it is seen that there is not just one event at any one point. For this reason, it would be wrong to be concerned with only one point in determining Hot Spots. A lot of factors that have caused an accident may have been effective before the accident occurred. If any roads are improved with the help of the maps, segments (±500 m) of accident points must be shown on the maps. Graduated segments, in particular, can show the way at the beginning and end of points of improving of risky segments by degrees. So, in this study, linear clustering, as a new technique, was used to determine the hazardous segments to prevent of traffic accidents and take precautions on the highways. Specifically, the results of risk assessment of 1 km segments are illustrated in linear clusters.

Assessment of Hot Pieces (HPCS) would be a more precise guideline for the study of risk factors and accident regions together in linear clusters, as seen in Fig. 3. By the same method, different maps, which show deaths, injuries and financial damage distributions, can be produced.

3.3. Illustration of potential Hot Spots (Probable Hot Spots)

Up to now, dangerous points have been shown as Hot Spots by points which exceed the upper limit. But when accident data are investigated, there are some points which will be on the Hot Spot limit after 1–2 accidents. Investigating these points is as important as for normal Hot Spot points. There are some points which are not Hot Spots but they will be Hot Spots after just one accident.

In this study, another important subject is taking preventative measures before a point becomes a Hot Spot.

Points that are not Hot Spots yet but have high-potential for being Hot Spots have not been mentioned in any applications so far. This study, it must be emphasized, includes segments with high-potential (95%) for being Hot Spots. In Fig. 4, Probable Hot Spots (PRHS) were determined and illustrated for the region approaching 1.645 value, at a rate of 95%. These probable Hot Spots can be compared with Fig. 2a–d. On routes 5 and 6 particularly,

Fig. 3. HPCS illustrations in linear clusters.

Fig. 4. Probable Hot Spots (PRHS).

some points are not Hot Spots but they will be soon. Urgent precautions must be taken for these points, therefore.

4. Conclusions and future work

Interpretation of results on a map is one of the fundamental guides. In this study, the map user is supplied with different visualizations of linear analysis illustrated with different techniques.

As a consequence of the application, however the hotspots have been evaluated by the number of accidents so far, studying with only accident numbers and mapping of them alone are not sufficient. Thematic maps show graduated accident parameters are the most important products for the study. It is important to determine which types of traffic accidents occur at which segments. The maps show the user not only kilometre values of accident points but also hazardous road segments. The number of people killed or injured, and the number of accidents with only financial damage, must be taken as information clusters individually and then HPHSs must be assessed through inter-sectional clusters of individual information clusters.

When route maps producing different criteria are investigated, as seen in Fig. 2a–d, Hot Spots appear at different points on the same routes. For instance, Hot Spots determined by mortality criteria are seen on straight routes because of the high speed on those sections. If there were an accident on this route, someone would normally be killed. Although the number of accidents with injuries is fewer than mortalities, the factors are more or less the same. Maps of damage Hot Spots show them appearing on international roads, which attract expensive vehicles. There is less mortality but more injured people and more accidental damage on international roads. These facts may give an idea about using goals and type of the routes and kinds of vehicles. However, the idea is not acceptable for the entire route. Nevertheless, the method should be applicable in other similar road networks especially using different accident parameters and considering all segments of roads.

In pieces maps, some segments do not have any Hot Spots, but some segments meet all the criteria. In this case, Highest Potential Hot Spots, which are determined by the intersection of all criteria must be considered first. Hot Spot maps producing different criteria are the most important tools for road improvement.

When we consider that the place where the accident happened is recorded with 1 km proximity, and the factors affecting the accident do not start just at the accident point, it is obvious that analysis of HPCS is critical. Table 1 shows the number of Hot Spots and priority status for all points, as another useful result for improving route planning.

When the results of the study compared with existing accident data, some useful suggestions can be mentioned. HPHS and PRHS numbers and locations which are determined by this method are different from normal Hot Spot numbers or locations. Naturally priorities of precaution types will be various at different points or road segments.

It can be seen that 22 high-potential Hot Spots could be real Hot Spots in the near future. These segments ought to be dealt with as a priority, in terms of preventative measures. All dangerous zones and the severity of accidents at the point of different criteria can be shown numerically from Table 1. So, both values and kilometres from maps must be used to improve current routes and planning new routes. The priority of the plan is reduction of mortality. It is also structurally problematic segments or points on routes that can be determined easily with the help of result values and terrestrial research. As an example, although route 5 does not have any

Table 1 Critical pieces and their values (for seven routes).

Reference	Hot Spot numbers	Priority status
Number of accidents	38	Hot Spots estimated from data
Number of deaths	37	1st priority spots
Number of injuries	27	2nd priority spots
Financial damage	27	3rd priority spots
HPHS	04	Most dangerous spots (to be analysed initially)
HPCS	38	Accident zones
PRHS	22	Highly potential hotspots necessitating precautionary measures

Hot Spots determined for damage, it has the most number of deaths, since it is so straight and crowded.

Investigating the dangerous spots and the road segments illustrated in the analysis maps prepared here, studying the contributory factors and taking precautions are the responsibility of the related administration.

References

Anselin, L., 1995. The local indicators of spatial association-LISA. Geographical Analysis 27 (2), 93–115.
Arslan, F., 2007. Emniyet Genel Müdür Yardımcısı. <http://www.trafik.gov.tr/icerik/ bildiriler/A6-87.doc> (accessed May 2008).
Berhanu, G., 2004. Models relating traffic safety with road environment and traffic flows on arterial roads in Addis Ababa. Accident Analysis and Prevention 36 (5), 697–704.
Berry, B., Marble, D., 1968. Spatial Analysis. Prentice Hall, Englewood Cliffs.
Block, R., Block, C., 1995. Space, place and crime: hot spot areas and hot place of liquor-related crime. Crime and Place 4 (2), 145–184.
Caliendo, C., Guida, M., Parisi, A., 2007. A crash-prediction model for multilane roads. Accident Analysis and Prevention 39 (4), 657–670.
Cho, W.K.T., 2003. Contagion effects and ethnic contribution networks. American Journal of Political Science 47 (2), 368–387.
Cliff, A.D., Ord, J.K., 1973. Spatial Autocorrelation. Pion, London.
Cliff, A.D., Ord, J.K., 1981. Spatial Processes. Models and Application. Pion, London. Ding, J.M., Pei, Y.L. 2007. Identification of road black spots and safety evaluation. In: International Conference on Construction and Real Estate Management, Bristol, pp. 21–22.
Eck, J.E., Weisburd, D., 1995. Crime places in crime theory. Crime and Place 4, 1–33. Faaliyetlerimiz Ve Hedeflerimiz. 2007. Turkish Ministry of Public Works and Settlement Turkey.
Gatrell, A.C., Bailley, T.C., Diggle, P., Rowlingson, B.S., 1996. Spatial point pattern analysis and its application in Geographical epidemiology. Transactions of the Institute of British Geographers 21 (2), 256–274.
Getis, A., Ord, J.K., 1996. Local spatial statistics. An overview. In: Longley, P., Batty, M. (Eds.), Spatial Analysis: Modeling in a GIS Environment. Wiley, New York.
Hilakivi, I., Veilahti, J., Asplund, P., Sinivuo, J., Laitinen, L., Koskenvuo, K., 1989. A 16 factor personality test for predicting automobile driving accidents of young drivers. Accident Analysis and Prevention 21 (5), 413–418.
Hu, J.B., Fei, X.L., Liu, X.M., 2007. Statistical method for identifying dangerous location and countermeasures. Journal of Beijing University of Technology 33 (5), 465–469.
Khan, G., Qin, X., Noyce, A.D., 2005. Wisconsin Traffic Operations and Safety Laboratory. http://www.topslab.wisc.edu (accessed June 2008).
Kristen, L.S.B., Mark, R.V., 2002. Improving the identification, analysis and correction of high-crash locations. ITE Journal 72 (1), 38–43.
Li, W., Carriquiry, A., Pawlovich, M., Welch, T., 2008. The choice of statistical models in road safety countermeasure effectiveness studies in Iowa. Accident Analysis and Prevention 40 (4), 1531–1542.
Ma, J., Kockelman, K.M., Damıen, P.A., 2008. Multivariate Poisson-lognormal regression model for prediction of crash counts by severity, using Bayesian methods. Accident Analysis and Prevention 40 (3), 964–975.
Maher, M.J., Summersgill, J., 1996. A comprehensive methodology for the fitting of predictive accident models. Accident Analysis and Prevention 28 (3), 281– 296.
Mccullagh, J.M., 2006. Detecting Hotspots in Time and Space. ISG06, University of Nottingham.
Meel, B.L., Safr, M.J., 2008. Fatal road traffic accidents in the Mthatha area of South Africa. South African Medical Journal 98 (9), 716–719.
Miaou, S.P., 1994. The relationship between truck accidents and geometric design of road sections: Poisson versus negative binomial regressions. Accident Analysis and Prevention 26 (4), 471–482.
Miaou, S.P., Lum, H., 1993. Modelling vehicle accidents and highway geometric design relationships. Accident Analysis and Prevention 25 (6), 689–709.
Mitra, S., Washington, S., 2007. On the nature of over-dispersion in motor vehicle crash prediction models. Accident Analysis and Prevention 39 (3), 459–468.
Moran, P., 1948. The interpretation of statistical maps. Journal of Royal Statistical Society 1 (2), 243–251.
Ng, C.H., Wong, Y.D., Lum, K.M., 1995. The applicability of the Poisson distribution in traffic accident count analysis – the Singapore experience. IES Journal 32 (2).
Nyerges, T.L., 1990. Locational referencing and highway segmentation in a geographic information system. ITE Journal 60 (3), 27–31.
Ord, J.K., Getis, A., 1995. Local spatial autocorrelation statistics: distribution issues and an application. Geographical Analysis 27 (4), 286–306.
Quddus, M.A., 2008. Time series count data models: an empirical application to traffic accidents. Accident Analysis and Prevention 40 (5), 1732–1741.
Ratcliffe, J.H., Mccullagh, M.J., 1999. Hotbeds of crime and the search for spatial accuracy. Journal of Geographical Systems 1 (4), 385–398.
Rathinam, C., Nair, N., Gupta, A., Joshi, S., Bansal, S., 2007. Self-reported motorcycle riding behaviour among school children in India. Accident Analysis and Prevention 39 (2), 334–339.
Rutter, D.R., Quine, L., 1996. Age and experience in motorcycling safety. Accident Analysis and Prevention 28 (1), 15–21.
Sherman, L.W., Gartin, P.R., Buerger, M.E., 1989. Hot Spots of predatory crime: routine activities and the criminology of place. Criminology 27 (1), 27–55.
Taylor, R.B., Gottfredson, S.D., Brower, S., 1984. Block crime and fear: defensible space, local social ties, and territorial functioning. Journal of Research in Crime and Delinquency 21 (4), 303–310.
Turkish Ministry of PublicWorks and settlement, 2007. <http://www.bayindirlik. gov.tr/turkce/dosya/2003\_2006faaliyetvehedefler.pdf> (accessed October 2009).
Vonderohe, A., Chou, C., 1997. A Generic Data Model for Linear Referencing Systems. Transportation Research Board, Washington (DC).
Weisburd, D., Green, L., 1994. Defining the street level drug market. In: Mackenzie, D.L., UchidA, C. (Eds.), Drugs and Crime: Evaluating Public Policy Initiatives. Sage, Thousand Oaks, CA.
World Health Organisation World report on road traffic injury prevention: summary, 2004. <http://www.who.int/world-health-day/2004/infomaterials/ worldreport/en/summary\_en\_rev.pdf> (accessed May 2008).

gundogduapplyinglinearanalysismethods2010.md 32 KB История Исходник