Analysis of the Risky Attitude of Driver Candidates by Graphical Models

ALATOOACADEMIC STUDIES; ISSN:16945263

Volume 5 Number 2 Year 2010; P:14-23

http://www.iaau.edu.kg/aas/AAS5-2.pdf

1

Analysis of the risky attitude of driver candidates by graphical models

VEYSEL YILMAZ a,*

, H.ERAY ELK a

CENGZ AKTA

a Eskiehir Osmangazi University, Faculty of Arts and Science, Department of Statistics, 26480

Eskiehir, Turkey

* Corresponding author. Tel.: + 90 222 2393750; fax: +90 222 2393578.

E-mail address: [email protected] ; [email protected]; [email protected];

[email protected]

Abstract

There are some good scientific studies conducted over risk-taking and risky behaviors of

reckless drivers; however, there are only a few studies related to attitudes displayed by

candidate drivers towards the traffic. The risky attitude of drivers is usually established by

determining their intentions about speed, safety belt use, and drunk driving. In this study, the

attitudes of the driver candidates towards drunk drivers and the use of safety belt in downtown

traffic were investigated from the perspective of gender, age, and educational level by using

graphical models. Graphical models provide a flexible tool for representing complex relations

among variables. These relations are marginal or conditional independencies and directed or

symmetric association. Each variable in a chain graph is represented by a node and some pairs

of nodes are connected by edges which indicate dependencies of subject-matter interest

whenever it represents a substantive research hypothesis. According to the results of statistical

analysis, while female driver candidates have a risky attitude regarding safety belt use and male

driver candidates have a risky attitude regarding drunk driving.

Keywords: Risky driving, Driver candidates, Graphical models
mailto:[email protected]:[email protected]




2

INTRODUCTION

In the years of 2005, 2006 and 2007; 570025, 664540 and 749456 traffic accidents occurred

respectively in Turkey and following numbers were recorded as casualty traffic accidents out of

these accidents 3195, 3365, and 3459 respectively. In Turkey, 5400 people were killed and

188,383 were injured in road traffic accidents in 2007. In Turkey, 502746, 680384 and 419137

individuals obtained drivers licenses in 2003, 2004, and the first six months of 2005,

respectively. According to the first six month data from 2005, 16561564 people have drivers

licenses. Considering that on average, 500,000 trainees a year work towards earning a drivers

license, it is a necessity to equip these candidate drivers with accurate cultural attitudes and

knowledge regarding traffic. Total highway network of Turkey is 63899 km. Density of Total

Network in meters per sq . Km is 82 meters. The number of vehicles being used is 12227393.

When we considered the population of Turkey, we may say that Number of Cars per 1000

Persons is 84 and Number of Cars per 100 Family is 34. On average, around 9000 accidents

occur in Turkey every year due to drunk-driving. Accidents associated with drunk drivers make

up 15% of all accidents. The death rate is the highest for drunk drivers at the ages ranging from

18 to 24 in all accidents happening in Turkey. This shows a parallelism to the traffic accidents

caused by speeding. Traffic problem of Turkey annually on average results in loss of 8-9 billion

dollars for country economy. When we compare Turkey with other countries in terms of traffic

accidents, we see that the result requires taking more urgent measures. While the death rate in

accidents per 100 million vehicle-kilometers is 0.9, 1.1, and 1.6 respectively, such rate is 20 in

Turkey (Traffic Education & Research Center, Report 2005-2007).

The population in Turkey increases gradually, but there is a considerable amount of migration

from the countryside to the towns because of health, education, and economics related issues.

This influx results in traffic problems that grow day by day. An established traffic culture does

not exist in the Turkish society and traffic education is not provided with an emphasis to close

the educational gap caused by the high rates of migration. However, there is a common belief

that such an education should be school-focused, rather than being a lifetime process. The

whole community should be aware of the need for traffic education, and the necessary




3

measures should be taken to determine drivers attitudes before they hit the roads, and to

remove the generally accepted negative tendencies and attitudes.

Studies mentioned above are conducted on active drivers. Traffic education and teaching a

traffic culture for candidates before they go out to traffic must be evaluated within a complete

framework. The attitudes and the behaviors of drivers do not come into existence as soon as

they go into the traffic. The subject carries his/her positive or negative attitude, obtained through

socializing, to traffic modified by the traffic education they receive. The knowledge of the

attitudes of drivers about these issues before becoming an active driver can provide an

opportunity to build an education system. In this study, the attitudes of driver candidates

towards the drunk driving and the use of safety belt in downtown traffic were investigated from

the perspective of gender, age, and educational level by using graphical models.

Attitude is an important factor determining the behavior of an individual .According to Allport

(1967) attitude is formed as a result of experiences in life and is an emotional and intelligent

readiness that has got a dynamic or diverting power to affect the behavior of the individual

towards all objects and situations. Attitude has an active role in the formation and development

of an action. Also, during their developments, acts may have roles that change the attitudes,

which form themselves. It has been thought that attitudes form a system that facilitates the

harmony of the subjects to their environment and have a secret directing power. One of the

most important terms about driver behavior is the term attitude. Driving styles reflect the

personal characteristics, attitudes, and motives of the driver (Elander et al., 1993). Before

getting started with changing the behavior of drivers, their attitudes must be investigated as a

first step (Forward, 1994). Some investigators in the traffic psychology field consider attitudes

involve personal characteristics that arise from a wide spectrum of motivators, including

aggression and risk taking; others investigate specific driver attitudes such as drunk driving or

speedy driving. On the other hand, some researchers suggest that attitudes should be

considered in the wider context of social environment or lifestyle (Aberg, 1996).




4

Most recent studies investigating how and when attitudes affect behaviors have been influenced

by Ajzens Theory of Planned Behavior (TPB) (1988, 1991). This socio psychological theory

offers a model of the relationship among believes attitudes, intentions, and behaviors (Parker et

al., 1998). Topics such as violation of traffic rules, drunken driving, attitude measurement, and

perceived risk have been investigated in the framework of Ajzens TBP. The human factor in

traffic was investigated in several scientific studies and drivers attitudes about traffic were

clearly shown. The accident tendency, personality predisposed to accidents, risk taking

tendency, and the driver behavior and attitude were investigated in detail (Parker et al., 1992,

1995a, 1995b, 1998; Summala, 1996; Yagil, 1998; Iversen and Rundmo, 2002; Ulleberg and

Rundmo, 2003; Elliott et al., 2003; Iversen, 2004; 2008). These studies had similar findings in

which the human factor has the highest percentage in accident formation and is found to be the

most difficult factor to be changed among the three factors (vehicles, human, and environment)

of the traffic system. Assum (1997) hypothesized that there are no direct relationships between

institutional attitudes and accidents and found out that drivers with correct attitudes about traffic

rules have lower accident risks when compared to those of drivers with bad attitudes. Based on

other studies, it is expected that young male drivers are more likely to violate traffic rules and

show more risky driving behaviors.

The tendency of drivers to have a risky attitude is usually determined by finding out their

attitudes towards fast driving, failure to use a safety belt, and drunk driving. Shinar (1993)

searched for relations between safety belt use and demographic, socio-economic factors. A

positive relationship was found between race, marital status, child number, education, income,

and safety belt usage. Stewart (1993) also investigated safety belt usage within the

determinants of accident and compared the driver behaviors before and after safety belt usage

act was released. Li, Kim and Nitz (1999) determined the same relation between safety belt use

among drivers and passengers. Moreover, in this study, strong relations were found between

alcohol and safety belt usage. Shin et al. (1999) investigated the relationships between safety

belt use and the socio-economic status and the ethnicity among high school students and

concluded that younger respondents view the usage of safety belts, in downtown traffic,




5

useless. Schechtman et al. (1999) studied the relation between alcohol drinking habit, speed

passion, and safety belt usage and found that alcohol passion is an indicator of risk taking in

traffic. They further discussed the relation between this habit and safety belt usage. Everett et

al. (2001), researched the security and risk behaviors of high school students in the USA

between 1991 and 1997. The results of this study revealed that 17% of the drivers drive after

alcohol intake and 33% do not use a safety belt. Calisir and Letho (2002) argued that safety belt

use was affected basically by personal factors such as age, gender, and race. Kim & Kim

(2003) in their research on safety belt usage, personal factors and severe accident injury,

showed that the ones who did not use safety belts have more severe injuries or deaths in traffic

accidents when compared to those of who use safety belts. They also found relations between

safety belt use and age, gender vehicle type, alcohol use, and speed. Chaudhary et al. (2004)

found that 29% of the American drivers did not use safety belts; people with high income levels

comprehended not using safety belts as a high risk, however, youngsters found it low risk.

In the current study, eleven hypotheses were investigated to find out the relation between

demographic characteristics and the risky attitude of drivers. Hypotheses are defined as follows:

H1. There is a relation between attitude regarding safety belt usage in downtown and gender.

H2. There is a relation between attitude regarding safety belt usage in downtown and age.

H3. There is a relation between attitude regarding safety belt usage in downtown and the

education level.

H4. There is a relation between the attitudes concerning driving under the influence of alcohol in

downtown and safety belt use.

H5. A There is a relation between attitudes on drunken driving and gender.

H6. There is a relation between attitudes on drunken driving and age.

H7. There is a relation between attitudes on drunken and the education level.

H8. Drunk driving and gender are conditional independent if age is given.

H9. Education level and safety belt use are conditional independent if is age given.




6

H10. Education level and drunk driving are conditional independent if the attitude on safety belt

use in downtown is given.

H11. Education level and safety belt use are conditional independent if the attitude on drunk

driving is given.

METHOD

Sample

The study sample consisted of 350 candidate drivers from a pool of 1530 who attended 17

driver courses, which are implemented under the Ministry of National Education Control, in

October 2004, in Eskisehir Twenty driver candidates selected randomly from each of 17 driving

courses constituted the sample of study. During the application of the measurement tool, 350

questionnaire forms were delivered to candidate drivers. Ninety two candidate drivers did not

fully or correctly answer and the statistical analyses were performed on the answers from the

remaining 258 candidate drivers. Table 1 gives the socio-demographic information about the

participants.

The variables and their levels were as follows:

a. Gender (male, female)

b. Age (18-24 years, 25-31 years, 32 years and older)

c. Education level (elementary school, high school, university)

d. I do not think that drunk driving is risky (I agree, I disagree).

e. I do not think that safety belt use is necessary in downtown traffic (I agree, I disagree).

Table . The demographical distribution of individuals in the sample

Statistical Procedure

In the study, Graphical Log Linear models, which provide an opportunity for numerical and

graphical interpretation, have been used (For the details and the proof for graphical model see:

Lauritzen, 1996; Pearl, 1993, 1995a, 1995b, 1998, 2000; Edwards and Kreiner, 1983;

Whittaker, 1990; Pigeot et al., 2000). Models that have interpretations in term of conditional




7

independence are known as graphical models. The terminology stems from the relationship of

these models to graph theory. Edwards and Kreiner (1983) give an overview of the use of

graphical log linear models.

Graphical models are a form of multivariate statistical analysis in which the structure of

dependences between variables can be displayed graphically. Note that graphs based on two

variable effects determine log linear model. Thus pictures of graphical models are worthless

until after graphical models have been defined. A key feature of this subject is the one-to-one

correspondence between graphical log linear models and graphs. Every model determines a

graph and every graph determines a model. The graphical models are represented by an

undirected independence graph. Graphs consist of vertices and edges as connections between

selected pairs of are used to formulate hypotheses about relation between variables. A graph

consists of vertices and edges. Vertices correspond to variables in log linear models. Edges

correspond to two variable effects. Vertices stand for variables connections representing

associations. When a missing connection is interpreted as a conditional independence, a graph

characterizes a conditional independence structure as well (Lauritzen and Wermuth 1989,

Wermuth and Lauritzen, 1990). For instance, according to the rules of notation, the graph of 4

independent vertices, shown in Figure 1, shows the shorthand notation [AB], [BCD] models,

M= U + UA + UB + UC + UD + UAB + UBC + UBD + UCD + UBCD (1)

where U is the general effect; UA, UB, UC and UD are the main effects ; UAB, UBC, UBD and UCD

are the two variable interactions and UBCD is the three variable interaction. From Figure 1, it can

be inferred. Given variable B, variable A is conditional independent of variables C and D. The

model shows shorthand conditional independence notation [A (C, D) B]. Graphical

models are determined by their two variable interactions. The basic idea is that any graphical

model containing all of the terms UAB , UAC and UBC must also include UABC . A model is




8

graphical if, whenever the model contains all two variable terms generated by a higher order

interaction, the model also contains the higher order interaction.

Figure 1 The independence graph of the log-linear model [AB], [BCD]

The best model may be selected either using a forward or a backward approach like in

regression analysis. In a forward selection approach, one starts with the independence model

and then progressively adds association parameters that significantly increase the fit of the

model. It first starts by testing all possible first order interactions; it then selects the one that

produces the most important and significant increase in fit. If it finds one, it is added to the

model. The other firstorder interactions are again tested, and terms are added if they are

significant. The procedure continues with higher order interactions until no other terms can be

added, yielding a final model that best fits the data. In backward selection, one starts from the

saturated model. It first tries to remove higher order interaction terms; if a term is removed, it

does not yield a significant difference. The procedure is repeated until the removal of a term

yields a significant difference. If the test with eliminating an edge comes to a solution like p,

since we cannot refuse the edge, there wont be any testing in the further iterations. (Yilmaz et

al., 2005).

RESULTS

The Eliminate Backwards method was used to reach a suitable model and to test the

hypotheses. The reason for this is the availability of a saturated model in the initial analysis. For

that reason, a saturated model was used for the first model. The result of the first step is given

below in Table II.

First step:

Table II. The first step results of the analysis done to reach the most proper model

As seen in the table, the edge with the highest p value and lowest chi-square value is ad edge.

For this reason, ad edge is removed from the saturated model. At the end of the first step,

[bcde],[ abce] preliminary model was reached. The graphic of the model is shown below:

Figure 2. The [bcde],[ abce] model graph




9

The second step: ce: 2 =14.8799 (df=12;p=0.2481). The ac and ae edges were left in the

second step. Since P value of ac edge was found to be higher than the level of significance, this

edge is removed from the preliminary models [bcde], [abce]. Subsequently, second preliminary

model would be [bcde],[ abe]. The graphic related to this model is given below:

Figure 3. The [bcde] ,[abe] model graph

The second step: ac: 2 =18.8850 (df=12;p=0. 0.0913) and ae: 2 =16.3964 (df=8;p=0.0370).

Only the ce edge was left in the third step. Since P value of the ce edge was found to be higher

than the level of significance, this edge is removed from the preliminary models [bcde], [abe].

Subsequently, the last model would be [bde], [bce],[abe]. This model can be expressed by

conditional liberty terms as [ ebda ,/ ] [ dbec ,/ ]. The graphic related to this model is

given below:

Figure 4. The [bde], [bcd],[ abe] model graph

Table III. The results of the hypotheses

The hypotheses, the results, and the graphics are shown in Table III. All hypotheses except H3

and H5 were accepted. No significant relation between education level and safety belt use and

drunk driving was found. However, relations between the variables could be determined

conditionally if the levels of the age were known. Besides, if the attitude about safety belt use is

known, the conditional relation between gender and alcohol was determined. Sixty four percent

of the men reported that they can drive while they are drunk. This rate was 54.9% in the

women. Odds ratio was found to be 0.66 between these two variables. This means that there

are 66 male candidates who consider alcohol as a risk with respect to 100 women who do so.

The only independent variable that is related to alcohol and safety belt use was found to be the

age variable. 60.3% of the subjects between18-24 ages regarded drunk driving as risky, on the

other hand 58.8% of the subjects between 25-31 ages regarded drunk driving as risky. 34.8% of

the subjects older than 32 years reported drunken driving as a risk. The odds ratio of attitude to

alcohol between 18-24 and 25-31 ages was found to be 0.97. The odds ratio between 18-24

and 32 + ages and 25-31 and 32 + was found to be 1.23 and 1.34, respectively. From this point




10

of view, the attitude about alcohol is similar between 18-24 and 25-31 ages, but the 32+ age

group do not see drunk driving as a risk.

A significant relation was found between gender and safety belt use. 65.5% of the men

considered safety belt usage unnecessary in downtown while 77% of women found it

unnecessary. The odds ratio of this statement was 1.76. This statement can be interpreted as

there are 176 women thinking safety belt use unnecessary while there are 100 men thinking it

necessary. The odds ratio for safety belt use attitude was found to be 0.7, 10.7, and 15.2 for the

age groups 18-24 and 25-31, 18-24 and 32 +, 25-31 and 32+, respectively. Especially, the 32+

age group has more attitudes about the unnecessary of safety belt use in downtown when

compared to other groups. The percentages of the subjects that declare safety belt use

unnecessary were 69% for 18-24 years, 62.7% for 25-31 years, and 60.5% for 32+ years.

DISCUSSION

In the current study, men of 32+ years of age and women who attend a driving course are found

to be risky driver candidates. While women have worse attitudes than men in safety belt use,

men tend to drink and drive more frequently than women. If safety belt use is accepted as a

personal flaw, it can be considered that they endanger only themselves. However, men who

drive under the influence of alcohol endanger not only themselves but also other drivers,

pedestrians and other people in the same vehicle as well. From this point of view, it is not wrong

to say that men have more dangerous attitudes than women. Using these results one can

extrapolate that men of 32+ years of age have dangerous attitudes in traffic. When the safety

belt use percentages are examined in Hawaii, Australia, Canada, USA, and Turkey as being

0.81, 0.91, 0.87, 0.58, and 0.21, respectively, it is very clear that the drivers in Turkey have bad

habits about this issue.

An individuals behaviors should be identified and understood through his/her attitudes. Like

many of the behaviors, the attitudes have also gained through learning. For these reasons,

before investigating drivers behaviors and attitudes in traffic, inferences may be made as to




11

they were involving an attitude toward to traffic before and during candidate driving process.

Individuals traffic education and teaching them traffic culture before they drive in the traffic

should be discussed in an integrated frame. Drivers behaviors and attitudes toward traffic do

not emerge immediately after they begin to drive in the traffic. While the individual is being

socialized, he/she carries his/her attitudes culturally gained toward traffic together with traffic

education gained as a response tendency in the traffic system which he/she actively takes

place. To determine candidate drivers attitudes on this subject before being active drivers

would provide opportunity to create an accurate training system or a process

The relation between demographic characteristics of the drivers and the risky attitude are

determined scientifically in the studies that exist in the literature. For this reason, these risky

attitudes should be determined during education and should be transformed into positive

attitudes bearing in mind that these candidates will be active drivers in the near future. It will be

useful to explain the harms of speeding, drunk driving, and the necessity of safety belt use with

the aid of visual communication devices and statistics in the lessons during the driving course

program.

REFERENCES

Allport, G.W. (1967) Attitudes, readings in attitude theory and measurement, Ed. Martin

Fishbein. New York: John Wiley & Sons, Inc, pp.1-14.

Aberg, L., (1996) The role of perceived risk of detection in the Theory of Planned Behavior.

International Conference on Traffic and Transport Psychology, 22-25 May, Valencia,

Spain.

Ajzen, I.,(1988) Attitudes, Personality and Behaviour. Milton Keynes: Open University Press.

Ajzen, I., (1991) The theory of planned behavior. Organizational, Behavior and Human Decision

Processes, Vol 50, pp.179 -204.

Assum, T. (1997) Attitudes and road accident risk. Accident Analysis and Prevention, Vol

(29),No.2, pp.153 159.




12

Calisir, F., Lehto M.R. (2002) Young drivers decision making and safety belt use. Accident

Analysis and Prevention, Vol 34, pp.793805.

Chaudhary, N.K., Salomon, M.G., Cosgrove, L.A. (2004) The relationships between perceived

risk of being ticketed and self-reported seat belt use. Journal of safety Research, Vol 35,

pp.383-390.

Driver Education in Turkey and World. T C General Directorate of Security Head of Traffic

Services Traffic Education and Research Center Report 2005-2007, Ankara, Turkey.

Edwards, D., Kreiner, S. (1983) The analysis of contingency tables by graphical models.

Biometrika, Vol 70, pp.553-562.

Elander, J., West, R., French, D., (1993) Behavioral Correlates of Individual Differences in

Road-Traffic Crash Risk: An Examination of Methods and Findings, Psychological

Bulletin, Vol 113, pp.179-294.

Elliott, M.A., Armitage, C.J., Baughan, C.J., (2003) Drivers compliance with speed limits: an

application of the Theory of Planned Behavior. Journal of Applied Psychology, Vol 885,

pp.964-972.

Everett, S.A., Shuls, R.A., Barrios, L.C., Sacks, J.J., Lowry, R., Oeltmann, J. (2001) Trends and

Subgroup differences in Transportation-Related Injury Risk and Safety behaviors Among

High School Students, 1991-1997. Journal of Adolescent Healthy, Vol 28, pp. 228-234.

Forward, S., (1994) Theoretical Models of Attitudes and the Prediction of Drivers Behaviour,

Uppsala Psychological Reports, No.434, Department of Psychology, Uppsala Unv.,

Sweden.

Iversen, H., Rundmo, T. (2002) Personality, risky driving, and accident involvement among

Norwegian drivers. Personality and Individual Differences, Vol 33, pp.1251-1263.

Iversen, H. (2004) Risk-taking attitudes and risky driving behavior. Transportation Research

Part F, Vol 7, No.3,135-150.

Kim, S., Kim K. (2003) Personal, temporal and spatial characteristics of seriously injured crash-

involved seat belt non-users in Hawaii. Accident Analysis and Prevention, Vol 35, pp.121

130.

Lauritzen, S.L. (1996) Graphical models. Oxford: Clarendon Press.




13

Lauritzen, S. and Wermuth, N. (1989) Graphical models for associations between variables,

some of which are qualitative and some quantitative. Annals of Statistics, Vol 17, pp.31-

57.

Li, L., Kim, K., Nitz, L. (1999) Predictors of safety belt use among crash-involved drivers and

front seat passengers: adjusting for over-reporting. Accident Analysis and Prevention, Vol

31, pp.631-638.

Parker, D., Manstead, A.S.R., Stradling, S., Reason, J.T. (1992) Determinants of intention to

commit driving violations. Accident Analysis and Prevention, Vol 24, pp.117131.

Parker, D., Reason, J.T., Manstead, A.S.R.,, Stradling, S.G., (1995a) Driving errors, driving

violations and accident involvement. Ergonomics, Vol 385, pp.1036-1048.

Parker, D., West R., Stradling, S., Manstead, A.S.R., (1995b) Behavioral Characteristic and

Involvement And Different Types of Traffic Accident. Accident Analysis and Prevention,

Vol 27,No.4, pp.571 - 581.

Parker, D., Lajunen T., Stradling, S. (1998) Attitudinal predictors of interpersonally aggressive

violations on the road. Transportation Research Part F, Vol 1, pp.11-24.

Pearl, J. (1993) Causal diagrams for empirical research. Biometrika, Vol 82, pp.669-710.

Pearl, J. (1995a) Graphical models, causality and intervention. Statistical Sciences, Vol. 8,

pp.266-269.

Pearl, J. (1995b) Causal inference from indirect experiments. Artificial Intelligence in Medicine,

Vol 7, pp.561-582.

Pearl, J. (1998) Graphs, causality, and structural equation models. Sociological Methods and

Research, Vol 27, pp. 226-284.

Pearl, J. (2000) Causality: models, reasoning, and inference. Cambridge: Cambridge University

Press.

Pigeot, I., Blauth, A., Bry, F. (2000) Interactive analysis of high-dimensional association

structures with graphical models. Metrika, Vol 51,No1, pp.53-65.

Schechtman, E., Shinar, D. , Comton R.C. (1999). The relationship between drinking habits

and safe driving behaviors. Transportation Research Part F, Vol 2, pp.15-26.




14

Shin, D. Hong, L. , Waldon, I. (1999) Possible causes of socioeconomic and ethnic differences

in seat belt use among high school students. Accident Analysis and Prevention, Vol 31,

pp.458-496.

Shinar, D. (1993) Demographic and socioeconomic correlates of safety belt use. Accident

Analysis and Prevention, Vol 25 6, pp.745-755.

Stewart J.R. (1993) Seat belt use and accident involvement: a comparison of driving behavior

before and after a seat belt law. Accident Analysis and Prevention, Vol.25,No6, pp.757-

763.

Summala, H. (1996) Accident risk and driver behavior. Safety Science, Vol.22, pp.103-112.

Ulleberg, P., Rundmo, T. 2003. Personality, attitudes and risk perception as predictors of risky

driving behavior among young drivers. Safety Science, Vol.41, pp.427-443.

Wermuth, N. and Lauritzen, S. (1990) On Substantive research hypotheses, conditional

ndependence graphs and graphical chain models. Journal of the Royal Statistics Society,

Series B, Vol.51, pp.21-72.

Whittaker, J. (1990) Graphical models in applied multivariate statistics, New York: John Wiley,

Sons.

Yagil, D. (1998) Gender and age-related differences in attitudes toward traffic laws and traffic

violations. Transportation Research Part F Vol.1, pp.123-135.

Yilmaz, V., Cangur, S., elik, H.E., (2005) Gender difference and earthquake experience effects

on earthquake victims. Personality and Individual Differences, Vol.39, pp.341-348.




15

Fig. 1 The independence graph of the log-linear model [AB], [BCD]

A B

C

D




16

Figure 2. The [bcde],[ abce] model graph

(d) (c)

(e) (b)

(a)




17

Figure 3. The [bcde] ,[abe] model graph

(d) (c)

(e) (b)

(a)




18

Figure 4. The [bde], [bcd],[ abe] model graph

,

(d) (c)

(e) (b)

(a)




19

Table I. The demographical distribution of individuals in the sample

Variable n %

Age

1824 184 71.3

2531 51 19.8

32 + 23 8.9

Gender

Male 145 56.2

Female 113 43.8

Education Level

Elementary school 71 27.5

High school 74 28.7

University 113 43.8




20

Table II. The first step results of the analysis done to reach the most proper model

Interaction 2

Degree of

free p

ab 37.0896 22 0.0231

ac 23.7370 18 0.1638

ad 14.9771 13 0.3088

ae 21.4303 13 0.0648

bc 149.8777 28 0.0000

bd 53.1245 19 0.0000

be 38.0345 19 0.0059

cd 36.6492 17 0.0038

ce 28.8983 19 0.0676

de 22.8647 12 0.0289




21




22

Table III. The Results of the hypotheses

Hypotheses Symbol Graphic Result

H1: ae a e

confirmed

H2: be b e

Not confirmed

H3: ce c e

Not confirmed

H4: de d e

confirmed

H5: ad a d

Not confirmed

H6: bd b d

confirmed

H7: cd c d

confirmed

H8: bda /

confirmed

H9: bec /

confirmed

H10: edc /

confirmed

H11: dec /

confirmed