ANALYZING THE CORRELATIONS AMONG DRIVER DISTRACTIONS AND THE PROBABILITY OF AN ACCIDENT

............................................................................................................................. ii Acknowledgement ............................................................................................................ iii The Tasks of this Master work ....................................................................................... iv Table of

iv

The Tasks of this Master work
Increasing mobilization has led to more and more people owning a car. The International Transport Forum of the OECD predicted that the number of cars worldwide will reach 2.5 billion by 2050. According to statista.com, 79 million new vehicles will be sold by the end of 2019. Driving has become a daily routine for many people. According to the U.S. Census Therefore, the focus of this work is deliberately not on the smartphone. The purpose of this master's thesis is to research the influence of one handed driving, background noise and the weather on the driving performance. The factors are deliberately chosen to correspond to a vehicle on a conceivable routine drive. It will be possible to make a statement with the results as to how likely driving errors are with these behaviors. After the identification of the test factors, a test plan using design of experiments was designed.
This assignment comprises of the following tasks: -Conduct a comprehensive literature review on the significance of distracted driving in road traffic and its impact on accident statistics to obtain overview of the current state of research on distracted driving.
-Design an experiment based on the results of the literature research. Within this framework, the main and blocking factors will be determined and the experimental design with the design of experiment will be created. Later on, a test environment vi on the driving simulator will be designed for the experiment and a survey will be created to determine the general driving behavior of each test group.
-Conduct the survey and the driving simulator experiment with participants who will make analysis possible.
-Evaluate the initial variables of the survey and the driving Simulator and evaluate them with a statistical model. The results will be written in a form of a thesis.
-Critical review of the developed approach and gained results -Oral presentation of the results    Table 17: Coefficient table of the experiment with modified

Introduction
The daily use of the car is an integral part of the everyday life of many people. Since the invention of the first automobile, the number of cars worldwide has increased every year.
In 2015, worldwide around 947 million passenger cars and 335 million commercial vehicles were in operation [1]. The International Transport Forum of the Organization for Economic Co-operation and Development (OECD) predicted that the number of cars worldwide will reach 2.5 billion by 2050 [2] [3].
The car is indispensable as a means of transport for most people, especially in regions with little or no local transport. Every American spends an average of more than 290 hours a year on the road, covering an average distance of 10,900 miles (17,541.84 kilometers) a year [4]. And by the end of 2019, around 79 million new automobiles are expected to be sold [5]. In line with the increasing number of cars, the number of accidents has also risen steadily in recent years. According to a report by the World Health Organization, more than 1.35 million people died in road accidents worldwide in 2018. In addition, another 50 million people were injured in road accidents in the same year. On average, 18 people per 100,000 inhabitants die each year in road traffic accidents worldwide. Globally, road accidents are the most common cause of death among children and young people between the ages of five and 29. Traffic accidents were the eighth most common cause of death among all age groups in 2016. More people die as a result of an accident than from tuberculosis or HIV [6].
The officially reported number of road deaths in the USA was 35,092 in 2016. Two years earlier, the number of road deaths was 32,744. This is 2,348 more than in 2014, and the number of road deaths per 100,000 inhabitants in the USA was 12.4 per year in 2016 [7].
Of these fatalities in 2016, around 34% were attributable to drivers or occupants of vehicles with four or more tires. Twenty-three percent are drivers or passengers of two to three wheeled vehicles. Three percent are bicycles and 22% pedestrians. Eighteen percent of accident fatalities cannot be assigned to a single category. The World Health Organization has compiled a list of the greatest risks in road traffic. These include speeding, driving under the influence of alcohol and other psychoactive substances, non-use of motorcycle helmets, safety belts, child restraint systems, unsafe road infrastructure, unsafe vehicles, inadequate care after an accident, inadequate enforcement of traffic rules and distracted driving [8]. Although vehicles are becoming safer, accidents are caused by driver carelessness or distraction.
In 2016, accidents attributed to driver distraction were 3,450, a decrease of 2.2% from the previous year. In 2015, the number of accident fatalities was 3,526. Nevertheless, 263 young people (aged 15 to 19) were killed by distracted driving in 2016. This corresponds to 10% of all fatalities in young motor vehicle accidents in 2016 [9]. The main causes of distraction while driving are telephone conversations, reading text messages, putting on personal cosmetics, searching for a place in the navigation menu or distraction caused by passengers [10].
The influencing factors that lead to distracted driving in Germany and the US will be evaluated in this thesis. According to the U.S. Census Bureau, the average driving time to work in the USA is 26.1 minutes. At 5 working days a week, the average commuter spends 4.35 hours a week and over 200 hours a year in the car [11]. For many people, daily driving has become routine. Especially, people for whom driving has become routine pay less attention to the actual process of driving. This is typical for frequent commuters.
In a study conducted by Allianzen Versicherung, a German insurance company, in 2011, 40% of respondents stated that they often use their mobile phones while driving. Far more than half eat or drink regularly while driving. More than 40% also adjust seat belts, seats and mirrors while driving. One third read or write text messages or e-mails while driving [10]. The smartphone has become an immense source of distraction, especially among young people. The desire to be constantly reachable and the fear of missing something has led many drivers to look at their smartphones while driving. This has led to many accidents in recent years. According to the U.S. National Safety Council, the use of mobile phones while driving encompasses 1.6 million accidents per year [9].
The distraction by the smartphone is a well-known distraction and has been addressed in various studies and publications. Therefore, the focus of this work is deliberately not on the smartphone. In the course of this thesis, the influence of one-handed driving, background noises and weather conditions on driving performance will be investigated.
The main factors are deliberately chosen to simulate the environment that prevails in many vehicles, especially in the morning. Listening to music from the radio is part of the normal soundscape of many drivers. In addition, some drivers consume a coffee or other beverages on their way to work. The danger of these everyday distractions, which have become a daily routine for many drivers and are no longer even perceived, is examined in more detail in this thesis.

Background
The vehicle population in both Germany and the USA has risen steadily since its introduction. This chapter explains the importance of the car and its current use in Germany and the USA. It also accounts for the different types of distractions and how they occur in road traffic. Based on the distractions, their influence on driving performance and the resulting accident risk is explained.

Cars and traffic in the United States and Germany
As the world population continues to grow, so does the number of road vehicles. Thus the volume of traffic on the roads rises. In almost all Western countries, the number of vehicles has risen continuously since the Second World War. The number of private cars has quintupled worldwide since 1960 [12] [13]. The USA and Germany were pioneers in motorway construction and promoted motorization at an early stage. Both countries have much in economic, political and cultural aspects in common. The economic upswing has led to a rising standard of living and growing per capita income in both countries. This has encouraged mobility, motorization and car use. The rapid expansion of the road network in both Germany and the USA since the Second World War has always made car use attractive and comfortable. In contrast to most other Western European countries, Germany has attempted to straighten the structure of the cities after the destruction caused by the Second World War and to orient them more towards car traffic. Urban planners often followed the example of American cities and the general growth of car traffic [12] [13].
Culturally, the car is a symbol of socioeconomic status, individual freedom and lifestyle in Germany and the USA [12]. Personal mobility in both countries is essentially based on the car. In the USA, there is a stronger one-sided dependence on the car, because many regions are sparsely or not at all connected to the local public transport network. Although Germany had similar growth in car ownership and use as the USA, it was more successful in limiting car traffic [13]. This is partly due to the preservation of local public transport.
For this reason, the following side effects of transport such as environmental damage, loss of landscape, health problems due to poor mobility, deaths and injuries due to traffic accidents, loss of productive time due to traffic jams, dependence on oil imports and the impairment of opportunities and accessibility for non-motorized population groups are not quite as pronounced in Germany as in the USA [12] [13] [14].
In the following section, the degree of mobilization of society in Germany and the USA will be examined in more detail.

Degree of motorization in Germany and the USA
Motorization in the USA and Germany has risen almost continuously over the past decades.
In 1939, there were 22 cars per 1,000 inhabitants. However, the Second World War reduced the number again, which is why in 1950 there were only 12 cars per 1,000 inhabitants [15] [16] [17]. Shortly before German reunification, the number of cars in West Germany had risen to 482 per 1,000 inhabitants. At that time West Germany was the country with the second highest motorization in the world, after the USA [18]. In East Germany, the degree of motorization was much lower. In 1988, there were 224 cars per 1,000 inhabitants. After the reunification of the two German states in 1990, the degree of motorization in united Germany increased by one fifth from 458 to 550 cars per 1,000 inhabitants by 2005 [19].
In the USA, the motorization of broad sections of the population began in the 1920s and 1930s, earlier than in any other country. By 1939, there were over 200 cars per 1,000 inhabitants, ten times higher than in Germany [20]. Figure 1 shows the development of motorization from 1960 to 1990 in Germany and the USA. Over time there has been a clear change in circumstances. In 1960, Americans owned four times as many cars as people in West Germany. In 1990, the gap in the degree of motorization had fallen to below 30%. Since 2000, however, a contrary trend can be observed again. By contrast, households with several cars occur twice as frequently in the USA (60%) as opposed to in Germany (27%) [22] [23] [24].   [26].
Although the gap in motorization between the two countries has narrowed over the decades (see Figure 1), the current level of motorization of German households with vehicles corresponds to that in the USA 30 years ago. This underlines the much earlier and stronger spread of car ownership in the US.
The next section describes the different use of cars in the two countries.  Figure 2 and Figure 3).  In contrast to the USA, however, other modes of transport continue to be used to a large extent in Germany. Of the remaining 40%, an average of 23% of journeys are made on foot, 9% by bicycle and 8% by public transport (see Figure 2). The national trends and shares in the choice of transport naturally hide differences between regions and cities [24].

Use of cars in the USA and Germany
Although the use of the car has increased in Germany, the figures show that the car is still used for a smaller proportion of journeys than was the case in the USA 30 years ago (see Figure 3). The following section describes the trends in car use in Germany and the USA. There are increasing trends in both countries in car use and time in a car as well as the number of hours spent in traffic. The increasing time spent in a car also increases the probability of an accident. Also, the probability of distracting oneself by secondary occupations increases due to the time in the car.

Traffic in Germany and the USA
In the following section, the distractions while driving will be discussed in more detail.

Inattention and distraction in traffic
Secondary activities in a car fundamentally affect the attention and responsiveness of drivers. Even if the distraction lasts a few seconds, the distraction provokes driving errors.
In several tests, the German ADAC Unfallforschung (ADAC Accident Research, equivalent to the US organization "AAA") has identified short inattentions of drivers due to a secondary activity while driving as the cause of an accident. Many experts regard the "distraction" factor to be just as dangerous as drunken driving [32].
Driving a car is a complex task that requires mental, physical, visual and acoustic skills from the driver, whether the car is on a busy road, in a city or on an abandoned country road. Concentrating on things other than the driving task increases the driver's risk of being involved in an accident. In 2012, the AAA Foundation for Traffic Safety conducted a road safety survey. The survey concluded that the majority of respondents see a great danger in "distracted driving". Of the participants 88.5% of the respondents considered making a phone call while driving to be a major risk. Nearly 95% of respondents felt that writing a text or checking email was also a serious security risk. Also, visiting social media accounts while driving is considered extremely risky by respondents [33].
However, many of the distractions to which drivers are exposed are not perceived by the driver as real distractions. According to the National Highway Traffic Safety Administration (NHTSA), the driver is considered to have disregarded the vehicle and is therefore at risk as soon as he performs an activity that requires his attention in addition to the main task of driving. The danger in the distractions generally lies in the fact that the analysis of the road for possible dangers does not take place. According to the AAA Foundation for Traffic Safety, typing the destination into the navigation system is one of the greatest safety risks when driving. On average, the driver is occupied with programming for about 40 seconds [34]. seconds required by the driver to adjust the navigation system. During this time, the driver does not look at the road, or only looks at the road occasionally, and will thus fail to recognize a possible change on the road later. The danger is that the less attentive the driver is to the road, then the less time the driver has to react to such a sudden change.
The Governors Highway Safety Association has divided the causes of driving distractions into four categories: visual, acoustic, manual and cognitive [35]. In the following sections, the four categories are explained in more detail. particularly conspicuous distractions such as work zones and certain advertising signs which are thus a risk to road safety [36].

Visual distraction
In another experiment, researchers Kircher and Ahlstrom used a driving simulator to investigate the influence of tunnel design and lighting on driving performance [37]. They concluded that as soon as the tunnel design and the lighting influence the visual attention of the driver, for example through color, the driver's performance decreases. However, the results also showed that bright tunnel walls can be used to direct the driver's visual attention forwards and thus increase safety in the tunnel [37].

Acoustic distraction
Acoustic distractions include activities that tempt the driver to concentrate on hearing or directing his gaze in a particular direction. Examples are conversations in a car or listening to music on the radio.
Researchers from the Department of Psychology at the University of Guelph, Canada, investigated the influence of playing audiobooks on driving performance [38]. They let the test persons drive in a driving simulator while the audiobook was running in the background and again in silence. While driving, the test persons were confronted with various situations, such as pedestrians or vehicles, which suddenly crossed the path of the vehicle. To measure driving performance, the researchers measured the braking reaction time to hazards. Surprisingly, the researchers came to the conclusion that playing an audiobook in a normal environment improves driving performance. However, as the environment became more complex, for example, due to bends, traffic or pedestrians, the audiobook contributed significantly to the driver's mental workload. In these environments, the addition of the secondary task led to a reduction in driving performance [38].
An important source of distraction in traffic is also telephoning while driving. The distraction of telephoning is not only attributed to the operation and holding of the device, but also, the division of attention between driving and talking is a distraction. For this reason, a hands-free system does not make it safer to make a call while driving than to make a call with a mobile phone in your hand.
Several studies have shown that hands-free and handheld telephony lead to the same high distraction of the driver. The researchers Desmet and Diependaele of the Knowledge Center Road Safety in Belgium have conducted a study on the effects of hands-free talking on eye movement patterns while driving in traffic [39]. For the experiment, thirty participants made two consecutive journeys of about 8.7 miles (14 kilometer) on a threelane motorway. During the first or second trip, each participant received a call on a handsfree device in the vehicle. The analysis of the eye movement pattern showed that other vehicles, traffic signs, or the speedometer were less fixed by the driver. The driver's visual scan pattern, on the other hand, showed a broader spatial distribution of eye fixations during a hands-free call. Based on these findings, the researchers concluded that the driver's eye behavior when making a phone call via a hands-free system is less strongly determined by the driving task. This means that the driver concentrates less on traffic-related information during the call and is more susceptible to not perceiving changes on the road directly, thus having a higher accident risk [39].

Manual distraction
Manual distractions include activities that require the driver to operate the vehicle in a compromise way that inhabits control. Activities such as adjusting the side mirror on a car, changing the radio station or typing the destination into the navigation system usually require the driver to take at least one hand off of the steering wheel. In the event of a change on the road, the driver may need to take both hands back to the steering wheel before reacting. In addition, manual distractions rarely occur alone when driving, but most distracting activities involve both a manual and another type of the four categories of distraction. An example of this is writing text messages manually while driving. For the driver, writing usually means a visual distraction in addition to manual distraction by the driver looking at the text field. The rapid spread of smartphones and social media has made the use of mobile phones one of the most significant manual distraction factors in road traffic in recent years. The constant use of the smartphone is part of the everyday life of many people.
The effect of writing text messages on the mobile phone while driving has been discussed in more detail by Motamedi and Wang (2016) in their paper "The impact of text driving on driving safety" [40]. For their experiment, they interviewed participants with a questionnaire about their behavior regarding typing while driving. They concluded that the majority of the participants are aware of the safety risks of typing. Nevertheless, more than a quarter of them reported frequently sending or reading text messages while driving.
Following the questionnaire, the researchers measured the respondents' driving performance in a driving simulator, while they were engaged in some forms of texting while driving and under different traffic conditions. As a result of the experiment, it was demonstrated that hands-free texting with a dictation function achieves much better driving performance in various challenging situations than manual texting with a mobile phone.
The authors also point out that the results of the study support the idea that reducing visual and manual distractions improves driving safety [40].
In another attempt to access distractions, a study on smartphones in road traffic has been carried out at the Highway Safety Research Center of the University of North Carolina.
The researchers in North Carolina investigated the "Effects of mobile phone distraction on pedestrians' crossing behavior and visual attention allocation at a signalized intersection" [41]. In an experiment, the researchers observed the behavior of pedestrians crossing a road. While crossing the street, the participants listened to music, made mobile phone calls or wrote a text message on their mobile phones. The researchers concluded that manual writing with a mobile phone as a secondary activity is the most demanding of our attention.
Pedestrians who were distracted by writing texts on their mobile phones looked around much less often when crossing the street. Also, they generally paid much less visual attention to their surroundings. Thus these pedestrians, who are distracted by this manual secondary activity, pose an immense risk to themselves and other road users through their inattentiveness [41].

Cognitive distraction
Cognitive distractions while driving include activities that distract the driver's mental attention from driving. They usually occur together with visual, manual or acoustic distractions, because these distractions cannot be mastered without the driver concentrating on the secondary activity.
Motamedi and Wang compared in their publication also the influence of Hand-Held and Hands-Free Cell Phone writing with the mobile phone [40]. In addition to writing a text message, they also tested the driver's response to reading a text message. They distinguished between those text messages that required an answer and those that should only be read. The evaluation clearly shows that the drivers who had read a text to which they did not have to respond with an answer showed significantly better performance when driving. It is likely that after reading the text, the driver will be able to focus his attention on the driving again, while the driver who is expected to answer is still mentally occupied with the answer [40].
Researchers from Tsinghua University (China), University of Leeds (UK) and Beihang University (China) also conducted an experiment to investigate the influence of cognitive distraction on driving [42]. The experiment aimed to understand how driving performance, measured here in terms of lane keeping power, changes during the simultaneous performance of a demanding purely cognitive task. The experiment was divided into three difficulty levels. While the participants drove in a driving simulator on a four-lane city street with a bicycle path and a sidewalk, they were told numbers. Depending on the level of difficulty, the numbers were asked to be recalled immediately, after some delay, or in a different order. The study comes to the conclusion that as the difficulty level of the task increases and the cognitive load increases, the lane keeping performance decreases noticeably. An increasing cognitive load has a negative effect on driving performance [42].
For a real driving environment, this means an increase in the probability of an accident. In the previous sections, the distractions were considered individually.
The next section looks more closely at the effects when different categories of distractions, such as visual, acoustic, manual and cognitive occur together.

Combination of different distraction
Divided attention is the ability of the human brain to perceive several different stimuli simultaneously. By dividing the attention, it is possible to process different sources of information and to successfully perform several tasks at the same time. This cognitive ability is important because it enables us to meet the demands of the environment, which almost always consists of more than one stimulus [43]. Cognitive processing is generally defined as "The mental action or process of acquiring knowledge and understanding through thought, experience, and the senses" [44]. However, the term cognitive is not used uniformly and describes a multitude of intellectual functions and processes. These functions and processes include information processing, knowledge formation, working memory, problem solving and decision making. Cognitive processes use existing knowledge and generate new knowledge.
The ability to share attention can be impaired by disease. Examples are schizophrenia or diseases such as attention deficit with hyperactivity disorder (ADHD). A craniocerebral trauma or a stroke can also lead to an attention deficit disorder. If the ability of divided attention is reduced, any distraction can affect the tasks performed simultaneously. For example, people affected by such impairments of divided attention find it harder to turn in a crossroads and speak at the same time. These people therefore have a higher risk of an accident in such situations [43].
The stimuli a person is exposed to add up to cognitive stress. Cognitive stress generally describes the mental strain to which a person is exposed [45]. If we consciously expose ourselves to two stimuli or activities at once and work with them, this is referred to by many as multitasking. Multitasking describes the ability to fulfil the requirements of several tasks simultaneously [46]. Multitasking is also called "continuous partial attention" in the English technical literature. It means the capacity of a person to receive simultaneous and possibly different stimuli [47].
Salvucci [48] describes multitasking as several (sub)tasks that a person integrates and performs in the context of a larger, more complex task. At the most basic level, multitasking can consist of actions that are directly related and require simultaneous execution. An example of this is changing lanes with the car, where attention must be drawn to the other lane and the steering wheel turned at the same time. On a more complex level, the tasks are either independent or nested and in both cases take place simultaneously. An example would be changing the radio station in the car while turning the steering wheel in a curve [48].
In their study "The effect of multitasking on the grade performance of business students", Ellis, Daniels and Jauregui [49] examined the impact of multitasking in more detail. The study examined whether multitasking in the classroom influences students' grade performance. To this end, the researchers experimented with 62 students of economics.
The students took part in a lecture for the experiment, then received a quiz on the content of the lecture. During the lecture, half of the participants were allowed to multitask in the form of texts with their mobile phones. The other half were concentrated completely on the lecture. The researchers came to the conclusion that the results of the students who were busy with texts were significantly lower than the results of the other student group. Based on this, the researchers concluded that multitasking during class had a significant impact on performance. The distraction of the students led to poorer performance in the tasks. A decisive factor for performance is the complexity of the task. In multitasking, the performance of people depends on the level of complexity of the task and the competence in each task [50]. When the performer is qualified in the task, he can perform these tasks with negligible impact on overall performance [50].
The more routinely an activity can be carried out, the greater the available knowledge and the easier it is to carry out other tasks. In addition, the perceived and sometimes real cognitive burden of a routine task is lower. Another example of this is reversing by car.
Many drivers listen to the radio while driving. But most people switch the radio off as soon as they have to park backwards. This is because the forward driving routine allows the driver to listen to the radio while driving. However, reversing is usually not a routine procedure and therefore means a higher cognitive load than driving forwards. Therefore, the driver turns off the radio and eliminates any secondary activity to minimize cognitive stress. In general, the effect of reducing cognitive stress through routine is helpful because it allows us to do several things at once. However, this becomes dangerous if the true cognitive load is misinterpreted. If too little attention is paid to an activity, the risk of errors is much higher.
If driving has become a daily routine for a driver, he can no longer be aware of the true cognitive burden of driving. This phenomenon also occurs with drivers who are supposedly familiar with the route they have travelled. In addition, the feeling of routine as well as driving on known routes leads to a rapid loss of concentration of the driver [51] [52].
In the next section, the different types of distractions in road traffic and their influence on driving performance, traffic and accident risk will be discussed in more detail.

Distractions and their influence on the driving performance
As described in section 2.2, the AAA Foundation for Traffic Safety concluded a survey in 2012 where the majority of respondents saw a great danger in "distracted driving". An even greater portion of 95.4% think reviewing or updating social media or using other functions of mobile phones (66%) while driving causes a higher chance of accidents [52]. However, the study also concludes that there is a significant discrepancy between participants' concerns and actual behavior in traffic. Although most respondents rejected distracted driving habits, the survey found that they practice many of these behaviors themselves.
More than one-third (34.6%) of respondents admitted to having read an email in the last 30 days while driving. One-quarter (26.6%) said they had written at least one text or email in the last 30 days. 68.9% of respondents had made at least one driving call in the last 30 days and almost a third (31.9%) admitted to doing so "quite often" or even "regularly". In addition, the NHTSA 2012 National Distracted Driving Attitudes and Behaviors Survey found that more than half (58%) of motorists surveyed continue their journey after receiving a call on their mobile phone [52] [53].
The impact of this behavior on road traffic is discussed in more detail in the next paragraph.  [55]. The researchers analyzed the data from Crashworthiness

Distraction as a cause of accident
Data System (CDS) from 1995-1999. The CDS records accidents that are so serious that a vehicle has to be towed away from the scene of the crime. The analysis of the data showed that 8.3% of drivers were distracted by crash-involved vehicles. In 36% of the accidents, however, it was no longer possible to determine whether the driver had left the vehicle. If the distribution of accident causes is applied to the 36% of accidents whose cause is known, the figure rises to 12.9%. However, because there is no evidence to support this assumption, the researchers have made a more conservative assumption. They assumed that the percentage of distraction among drivers with unknown attention status was only half that of distraction among drivers with known attention status. With this conservative approach, the total share of distraction among crash-induced drivers is still about 10.6% [55] [56].
The researchers also carried out an analysis of the CDS data for the years 2000-2003. It was found that only 6.6% of crash-involved drivers were distracted during this period.
However, the number of accidents in which the driver's attention status was unknown at the time of the accident was 46%. It applied the same conservative assumption based on the distribution of accident causes during this period. Thus, for the period 2000-2003, the total proportion of distracted drivers involved in an accident was 10.4%.
Thus over the period from 1995 to 2003, the proportion of accidents caused by distracted driving which were so serious that at least one vehicle was towed away from the scene of the accident did not noticeably increase. If it is assumed that the transfer to accidents with unknown distraction status is correct, then the percentage of deflected crash-induced accidents in this period is 10.5% [55].
In addition to the proportion of accidents caused by driver distraction, the study of Stutts, Reinfurt, Staplin and Rodgman [55] also examined in detail the cause of distraction and how it has changed over time. In the 1995-1999 analysis of reported accident data, researchers found that about 70% of reported distractions were inside the vehicle.
Passengers and audio equipment were the most common distractions. The remaining 30% were distractions outside the vehicle [55].  [57].
As described at the beginning, a great danger with many secondary activities is that the driver does not perceive them or no longer perceives them as distractions. In the following section, the distractions during routine journeys are examined in more detail.

Effect of distractions on driving behavior
In 2001, the Highway Safety Research Center at the University of North Carolina conducted an observational study to investigate how drivers behave in traffic. The study looked more closely at the types of activities that drivers perform while driving. For this purpose, the possible consequences of the activities on driving behavior were examined [56].
Seventy drivers have been selected for the study, who were recorded a week in their own vehicles on their normal journeys. About 10 hours of these journeys were recorded on video by the researchers. These videos have been analyzed to identify distracting side tasks and to analyze how often the driving activities have been carried out. The effects this has on traffic and by how much the accident risk increases as a result of secondary employment will be discussed in the next section.

Analysis of the distraction for the increase of the accident risk
The Virginia Tech Transportation Institute (VTTI) conducted the 100-Car Naturalistic Driving Study for the NHTSA [59]. The researchers aimed to calculate a quota ratio that represents the relative risk associated with a particular secondary task. One hundred drivers were selected to commute daily in or around the northern Virginia/Washington, DC area.
Each of these participants uses either their vehicle or a leased vehicle for their journey to work. The sample was limited to six vehicle types due to instrumentality issues. The group of participants had been chosen to include a disproportionate number of younger drivers between 18 and 25 years of age and drivers with high annual mileage. From other studies, these two groups were assigned the high accident probability, which was intended to maximize the potential for recording crashes and near-crash events for the experiment. In the course of the experiment, over a period of 12 to 13 months, depending on the participant, a total of over 2 million vehicle miles and around 43,000 hours of driving were recorded on video [59].
During this time 69 accidents, 761 near accidents, and about 20,000 baseline segments could be recorded. The baseline segments were randomly selected to represent normal, uneventful driving. During the analysis of the videos, the researchers came to the conclusion that 33% of the recorded accidents and 27% of the recorded near misses were caused by a secondary task [59]. Using the recorded near crash data and basic data, the researchers calculated the odds ratio. The odds ratio represents the relative risk associated with a secondary task. For this purpose, the researchers divided the secondary tasks into three categories based on the time the driver had to take his eyes off the road. Complex tasks required more than two keystrokes or glances off the road, thus increasing the risk of an accident. These activities include applying make-up and reaching for a moving object.
Moderate secondary tasks, on the other hand, are tasks with a maximum of two keystrokes, for example. These activities include inserting a CD and cassette or eating while driving.
The third category includes simple tasks that required at most a keystroke or a glance.
Activities that belong to this category are the stopping of radio, drinking or smoking [59].
The researchers calculated the quota ratios for the categories. This resulted in a quota ratio of 3.1 for complex secondary tasks, 2.1 for medium secondary tasks and 1.0 for simple secondary tasks. This means that when performing a complex ancillary task, drivers have about three times the risk of being involved in an accident or near miss as drivers who do not perform a category ancillary task. For medium secondary tasks, there was about twice the risk of driving without secondary tasks. For simple secondary tasks, there was no significant increase in the crash and near-crash risk [59].
The results of the quota ratio suggest that the complexity of the secondary task influences the crash and near-crash risk. Based on these findings, the researchers performed additional analyses to identify the environmental conditions associated with distraction-related crashes and near-crashes.
For these analyses, they only considered the complex secondary tasks with increased quota ratios, which indicate an increased risk. They concluded that quota ratios increase further when certain additional conditions occur together with a complex secondary task performed by the driver. These conditions, which increase the risk of accidents, include: twilight and unlit darkness, rain, divided roads and roads with gradients (straight or curved). The fact that normally divided roads are considered to be safer than undivided roads supports the conclusion that the risk of an accident is significantly increased by the driver performing a secondary task.
The studies described in this section underline the importance of considering distraction as a separate problem that differs significantly from other categories of carelessness. To investigate in more detail how the driver's behavior increases the probability of an accident, observational studies that are as realistic as possible are of particular interest.
This paper examines a combination of possible distraction factors. These distraction factors are deliberately chosen in such a way that they reflect an easily conceivable scenario for the test persons, because they are conceivable in one form or another in most vehicles. The results make it possible to determine the accident risk of commuters who travel the same distance every day. The following chapter describes the test setup and how to perform the test.

The experiment design
As described in section 2.2, distractions when driving can be divided into different categories. Each distraction has a different effect on the driver while driving. This chapter describes the experiment that resulted from the literature research and is analyzed in this thesis.

Statement of the problem
In both Germany and the United States, human error is by far the most frequent cause of road traffic accidents (Germany: 91% [60], USA: 94% [61]). Alcohol, drugs, fatigue, inappropriate speed and distraction are common causes and key factors of human misconduct in road traffic. Contrary to the technical standards in cars, the behavior of the driver and the road users, which seems to be difficult to influence or optimize, is the most important factor. For more safety in road traffic, however, it is essential to sensitize all road users to safe behavior. The German Road Safety Council summarized this in a statement in 2010 that in addition to the technical and legal measures, the assumption of responsibility by road users but also by CEOs and politicians was indispensable [10]. The cornerstones of such a system must be safe roads and roadsides, acceptable speeds, safe vehicles and informed road users. To achieve this, the system must be designed to forgive human error as far as possible.
The increasing use of the car (see 2.  [63]. Some driving activities that are not relevant to driving are already regarded as acceptable under common law by drivers and their danger is therefore no longer perceived. These include, for example, activities such as eating, smoking or selecting a radio station, which distract the driver from driving. For many commuters who drive a long distance to work every morning, it has often become routine to drink coffee or other beverages while driving.
This thesis analyzes the relationship between driver distractions and accident probability.
The null hypothesis is that the main factor level in Table 1 has no influence on the driver's driving performance. The alternative hypothesis is that the main factor levels in Table 1 influence driving performance. Then this study seeks to identify the influence of Age and Gender Blocking according to driving performance. Hence, the null hypothesis is that the blocking levels in Table 2 have no influence on the driver's driving performance. The alternative hypothesis is that the blocking factor levels in Table 2 influence driving performance. The survey and a driving simulation experiment will be conducted to investigate the effects of distracted driving behavior. It will be possible to determine the influence of a combination of everyday secondary activities on accident risk. Based on the results, a statement can be made about the extent to which certain driving behaviors increase the probability of driving errors.
In the next section, the procedure of the study will be described in general.

General description of the study
As explained in the introduction, there are studies that suggest that different forms of distraction and their combination have an impact on the driver's driving style. This thesis explores the magnitude of the negative effects of certain distraction factors and their combinations. The distraction factors that will be investigated in more detail are the background noise in the car, hand on wheel while driving, and different external weather conditions. The main factors are deliberately chosen so that they could occur in a large number of cars, as secondary occupations and circumstances on a daily commuter trip, for example. That is, a person who goes to work in all weather and drinks a to-go cup on the way. With the results, it will be possible to analyze the connection between driver distractions and to make statements about the accident probability of a driver who is exposed to these circumstances and secondary occupations.
In order to participate in this research project, respondents must be at least 18 years old.
The experiment begins by first questioning all participants in the experiment in a survey.
The survey examines the driving habits and behavior of the participants. The demographic data of the participants, such as age and gender, as well as a personal assessment of their own driving habits are asked. Subsequently, the participants take part in the second part of the experiment wherethe participants drive in a driving simulator. The experiment consists of 12 repetitions with the same driving environment in the simulator, which are performed in random sequence. In each repetition, a different combination of the above mentioned main factors will be performed.
A driving simulation aims to investigate the driving behavior of the individual with the distraction factors in different scenarios, such as braking events, signs and traffic rules. The experimental design allows a complete analysis of the performance of each participant under different types of conditions and different forms of distraction.

Participants
A total of 57 people participate in the survey and 20 finish the simulator experiment. Two groups are taken into account for the age of the participants and also for gender. Both

IRB approval
The approval of the Institutional Review Broad (IRB) at the University of Rhode Island has been obtained to conduct the experiment (IRB1819-161). The IRB ensures that the rights and welfare of recruited individuals for research projects are protected. For approval, the IRB has been provided with the objective of the research, research design and methods, the location of the research, the number of participants, ways of recruiting, security and confidentiality procedures to ensure that the approach of the study complies with federal regulations, guidelines and procedures.

Survey about driving behavior
The survey on the general driving behavior of the individual test groups is carried out using the SurveyMonkey Internet service. Appendix A provides a printed version of the survey with the declaration of consent that the participants had to agree to.
Before they begin the survey, the respondent is given a short explanation of the procedure and confirms in the consent form that they agree to the collection of their data and participation in the study.
In the first part of the survey, the respondent is asked to assign themselves to an age group (under 40 years; over 40 years) and gender (male, female). Then the driving experience of the participant is queried by how long he has had a driving license (1-5 years, 6-10 years, over 10 years), how much he drives (more than once a day, once a day, twice a week, once a week twice a month, once a month, other) and how long on average per day (less than 30 minutes, 30 -60 minutes, 1 -2 hour, 2 -4 hours, more than 4 hours). With reference to the later experiment in the driving simulator, each participant is then asked whether he is aware of any distracting behavior when driving. More precisely, he is asked whether he sometimes drives with only one hand on the steering wheel and whether he listens to music or the radio while driving. In the last question, the participant is generally asked whether he is aware of the fact that distracted driving increases the probability of an accident.
The survey later makes it possible to assign the participants to a certain group and to make general statements about the driving behavior of this group. With SurveyMonkey, the data could be downloaded directly into Excel for further analysis (see Appendix B).
The next section describes the test in the driving simulator in more detail.

The driving simulator experiment
The driving simulation experiment was developed to evaluate the effect of different forms of distraction while driving. In the following sections, the driving simulator itself and how the experiment was performed are described in more detail.

The driving simulator
A TranSim VS IV Driving Simulator (Model 03755) manufactured by L3 Corporation, as shown in Figure 4, is used for data acquisition. The vehicle controlled in the scenario can choose from a variety of predefined vehicle configurations. The simulator cockpit (see Figure 5) has a driver's seat, steering wheel, brake pedal and accelerator pedal. Each driver starts with 100% in Miscellaneous and 100% in Vehicle Handling Safety. The software assigns a certain percentage to each offense, which is subtracted from 100%. For example, High Speed Collision and Maximum Speed Limit Exceeded are rated at 5% each and Hard Braking at 3% and deducted from Vehicle Handling Safety. The driver in the experiment is not informed of the loss of points but can assume that the loss of points is equivalent to real traffic offenses. After the drive has been completed, the final value is given in the Driver Assessment as the driving performance value and output variable.

The experiment with the driving simulator
All experiments for the thesis were performed in the Driving Simulation Lab of the University of Rhode Island. With a separate program called Scenario Builder™, the scenario was developed and provided with the desired conditions. In order to carry out the experiment in such a way that it applies as universally as possible to many people who drive a lot, the scenario environment includes both main and secondary roads as well as highways (see Figure 6).  In the next section, the design of the experiment with the driving simulator is described in detail.

Design of experiment
In the thesis the effects of distracting behavior during driving are examined. The driving simulation experiment has been developed to evaluate different forms of distractions and their combination while driving. The participants drive in the simulation as a car in suburban traffic. The factors investigated in the developed driving simulation experiment were divided into main factors and blocking factors (see Table 3). A total of three main factors and two blocking factors are measured. As shown in Table 3, Hand-On-Wheel, Background-Noise and Weather are defined as the main factors.
In Hand-On-Wheel, one-handed driving is compared to two-handed driving. Second, it examines the extent to which Background-Noise has an influence on the driver's driving behavior. Finally, the Weather is the third main factor, and the influence of weather conditions on driving performance is investigated.
In the survey (see session 3.5), participants will be divided into under 40 (18-40 years) and over 40 (41-90 years) age groups and two genders. Based on this, the age groups and the two gender groups become two blocking factors (Age and Gender) in the experiment. In Based on the fact that the main factor Hand-On-Wheel has two levels, the main factor three Background-Noise has three levels and the main factor Weather has two levels, there are 12 combinations of the main factors (see Table 4).

Measurement and collection of data sets
While the participants drive in the driving simulator, after 4 minutes a repetition is over and the researcher stops the scenario in the driving simulator.  As can be seen in Figure 7, the violations are assigned to either the Vehicle Handling Safety or the Miscellaneous category. In both categories, each driver starts with 100%. The driving assessment is saved for each repetition. This means that 12 driving assessments are stored per participant, from which the driving performance can be determined by the researchers. The procedure to determine the driving performance are descripted in the next sections. Braking mean a deduction of 3% each in Vehicle Handling Safety. If the software detects a violation while driving, the percentage is deducted from the corresponding category. The percentage deduction is always the same for a particular violation. The final value is given in the Driver Assessment and is the automatically generated output variable of the driver's performance in the driving simulator (see Figure 8). In Appendix B is a complete driver assessment. The problem with this evaluation, however, is that the simulator's evaluation of the violations does not always appear appropriate. For example, a High Speed Collision is a deduction of 5% in the category Vehicle Handling Safety (see Figure 9). The same number of points will also be deducted for Maximum

Output factors of the driving simulator
Speed Limit Exceeded (see Figure 9). Miscellaneous is not used as a starting variable for this thesis.

Used output factor of the experiment
Instead of the percentage output variable in the driving assessment, a separate evaluation is defined for the experiment. The basis for the evaluation is the listed violations in the driving assessment. In this way, it is ensured that the trips continue to be evaluated neutrally and equally. It is assumed here that the driving simulator software always reports the same driving errors under the same circumstances.
In the course of the test, the offenses listed in Table 5 were detected by the simulator. Traffic Laws serve as the basis for this rating system. The proposed penalty is used as a distinguishing and weighting criterion. A legal sentence or the withdrawal of a driving license is considered to be more severe than a fine. Table 6 shows which paragraphs of the Rhode Island Traffic Laws govern the violation and what penalties they impose.

Depends on the Damage
From the penalty that the Rhode Island Traffic Laws sets for the violation, and the external circumstance, a penalty point number will be set for each violation. The penalty points form the evaluation system from which the new starting variable is calculated for each ride in the driving simulator.
First of all, it is determined for the evaluation that the same offenses receive more penalty points at higher speeds. On the other hand, this can be justified by the formulas for the braking distance (see Eq. (3)).
For the braking distance, the speed is the decisive factor for the distance until the car comes to a standstill. The higher the speed, the greater the risk that the car will not stop in time for a dangerous situation. Speed is also a decisive factor in a collision. The higher the speed of the car in a collision, the higher the force acting on it. The basis for this is the conservation of momentum theorem, which can be derived from Newton's Second and Third Axioms (see Eq. (4)).
The equation (4) shows that for a car with the same Mass ( ) and the same Time ( ), the acting force increases in an accident with increasing speed. The force acting in an accident determines how the car and the object hit are impacted and thus the extent of personal injury or damage to property.
Then a distinction is made between two categories when assessing infringements. The first category includes infringements that are dangerous but do not directly involve personal injury or damage to property. These include, for example, violations such as Hard Braking or Following too close. On the other hand, there are offenses that directly involve personal injury or property damage, such as Speed Collision or Cab collided with a structure.
According to the Rhode Island Traffic Laws, the penalty level for Maximum Speed Limit Exceeded is divided into two levels. First, less than 10 mph above the respective speed limit and 10 mph above the speed limit (see Table 6). Even if the fine for speeding is high, only the driver of the car is involved. If the offense is considered in isolation, the possibility of an accident is not certain. In addition, driving a maximum of 9 mph above the speed limit applies to most drivers. For this reason, the offense will be assessed with 10 penalty points. Each speed violation of more than 10 mph will be assessed with 30 penalty points because fees for this violations based on Rhode Island laws are between two and three times higher (according to RI Gen L § 31-41.1-4). On the other hand, the general risk is significantly higher with increasing speed.
The same argument also applies to driving Too fast for weather conditions. It means a driving style does not suit the conditions on the road even if it is possibly within the rules that are otherwise prescribed on the road. RI Gen L § 31-14-1 (2013) sets a penalty of $85 for this offense. Since this is also consistent with the penalty for the first stage of speeding, the same penalty points are used for this offense.
Following too close, unlike driving too fast, means that the driver is on the road with another vehicle. In addition, the possibility of an accident is increased by driving too close, which makes an accident more likely. The fine for Following too close according to RI Gen L §31-15-12 is however comparable to the fine for Maximum Speed Limit Exceeded  Table 7. The points for the offenses are summed for eachrun and form as a sum the initial variable of the experiment repetition. This initial variable is used to evaluate the experiment. In the next chapter, the experiment with the initial variable is evaluated and it is determined which main factors have the greatest influence on driving performance.

Results and analysis
This chapter presents the results of the survey and the experiment. After a description of the collected data, the collected data are statistically analyzed.

Results of the survey
The survey had 57 participants. The results of the survey are summarized in Table 8. License for more than 10 years.
In the next questions the participants were asked how often they drive a car. On this question the participants could choose between more than once a day, once a day, twice a week, once a week, twice a month, once a month and others. The answers were divided as shown in Figure 10. In the following question the participants were asked how long the participants drive on average per day. The participants could choose between less than 30 minutes, 30 -60 minutes, 1 -2 hours, 2 -4 hours and more than 4 hours. The distribution of answers is shown in Figure 11. In the last question, the participants were asked whether they were aware that distracted driving increases the probability of an accident. All respondents answered yes to this question.
The survey is a general statement about the participants in the experiment. Over 60% of the participants have had their driver's license longer than ten years and over 80% of the In summary, the overwhelming majority participants are experienced drivers with enough driving experience who, however, have the potential to be affected by the main factors of the experiment, such as one-handed driving, even in a real situation.

Results of the experiment
The In this section, the results are generally described in more detail. The raw data that is used in the following sections can be found in Appendix C.

Descriptive analysis for responses
This experiment with the driving simulator seeks to identify the three main factors with their levels: Hand-On-Wheel (One-Hand, Two-Hand), Background-Noise (Silent, Music, News), and Weather (Rain, Snow) that made 12 combinations of factor-level treatments.
A total of 20 participants completed the 12 repetitions in the driving simulator. The 12 replications which each participant made, lead to a total number of responses of 240 (see Table 9). The sample mean of the response is 37.31, the sample median is 30 and the sample standard deviation of the responses is 39.51. Based on the fact that the sample mean is slightly larger than the median, it can be concluded that the data distribution is positively skewed.

Check normality plot
The following Figure 13 shows the probability plot of the output variables of the experiment.

Figure 13: Probability plot of the responses [created with Minitab]
The normality test hypothesis assumed the null hypothesis H0: data are normally distributed versus the alternative hypothesis H1: data are not normally distributed. The normality test gives a p-value < 0.005 (see Figure 13). The p-value < 0.05 applies and the hypothesis H0 at 5% level of significance is rejected. This means that the data responses are not normally distributed.
In the following section, the responses are analyzed with the Analysis of Variance (ANOVA).   Figure 14 shows which of the main factors have the greatest influence on the responses. For the experiment of this thesis, the Pareto diagram shows that the main factor Background-Noise is the only one that has a significant influence on the driving performance (see Figure 14).

Figure 14: Pareto chart of the experiment [created with Minitab]
With the ANOVA, the input variables are examined more closely in order to check their significance and to identify the essential factors (main and interaction factors). All factors with a p-value at α=0.05 are considered significant for the output variable. Table 10 shows the ANOVA table of the experiment. The ANOVA analysis shows that based on the p-values in the ANOVA table, only the input variable Background-Noise with the p-value = 0.045 indicates statistical significance.
The blocking factor Gender has a p-value of 0.077 and is therefore marginally significant.
All other main factors and possible interaction factors are not statistically significant due to their p-value and will therefore not be investigated further.
With the Coefficient Table, the significance of the individual levels of the significant input factors can be examined more closely (see Table 11). Background-Noise is the significant factor for the responses with p-value is 0.013. The blocking factor Gender has the p-value=0.077. In the next section, the blocking factor Gender is analyzed.

Analysis of the blocking factor Gender and unmodified responses
In order to analyze more precisely whether certain main factors may have a particular significance for the blocking factor Gender, the responses are divided into two data sets. First, the Table 12 shows the descriptive statistics of the responses of the blocking factors Gender: Male.  Table 12 shows that the total number of responses is 120 and the Mean of this data set is 81.50. The range is from zero to 290 punishment points. The distributed responses of the data set Male is shown in Figure 15.

Figure 15: Chart of unmodified responses for the data set Male [created with Minitab]
Then the significant factors for the responses of this data set are examined. Figure 16 shows the Pareto diagram of the main factors, the blocking factor and the unmodified responses. The Pareto diagram gives an estimate of the relative importance of the main factors and interactions that have the greatest influence on the initial behavior of the data set Male. For this data set, the Pareto diagram shows that no factor has a significant influence on the driving behavior (see Figure 16). The ANOVA Table (see Table 13) of this data set also shows that no factors seem to have a significant impact on this group. Then, the significant factors for the second data set are analyzed. Table 14 shows the descriptive statistics of the responses of the data set Female. The data set Female also has a total number of 120 responses. The Mean of these responses is 66.92 and the range is from zero to 390 punishment points. The distribution of the responses are shown in Figure 17.  The unmodified responses of the data set Female are analyzed with the main factors, the interactions of the main factors and the blocking factor Age. The Pareto diagram shows that the main factor Weather is the only one that has a significant influence on driving behavior (see Figure 16). Table 15 shows the ANOVA Table of this analysis. The ANOVA Table also shows that the main factor Weather is the only significant factor with a p-value of 0.08.  Figure 19 shows the Pareto chart of standardized effects.

Figure 19: Pareto chart of the experiment with modified responses [created with Minitab]
Again, it shows that the Background-Noise is the most important factor influencing driving performance followed by interaction between Background-Noise and Hand-On-Wheel, and the interaction between Hand-On-Wheel and the Weather. The ANOVA shows the following p-values for main and interaction factors (see Table 16). The ANOVA table shows that Background-Noise has the highest significance for responses, with a p-value of 0.029, but also the interaction between Hand-On-Wheel and Background-Noise (p-value=0.032) and the interaction between Hand-On-Wheel and Weather (p-value=0.048) are statistically significant. In addition, the Age blocking is statistically significant with p-value=0.048.
However, the intensity with which the main factors and interactions relate depends on the level of input factors. The Coefficient Table can be used to more accurately analyze which levels of significant factors have the greatest influence (see Table 17).    (see Table 16), only the interaction between Background-Noise and Hand-On-Wheel and between Hand-On-Wheel and Weather are significant. Figure 20 shows the interaction between the Hand-On-Wheel levels (One-Hand and Two-Hands) and Background-Noise (Music, News and silence), and Weather (Rain, Snow). Figure 20 also shows that there is a significant difference in the interaction between the Hand-On-Wheel levels and the Background-Noise levels. Driving with one hand with Background-Noise will result in a higher number of penalty points than driving with two hands without

Background-Noise. The interaction between Hand-On-Wheel levels and Music as
Background-Noise has the highest number of penalty points while driving of all Background-Noise levels (see Figure 20).
The evaluation of the results shows that this interaction is not statistically significant with p-value = 0.481 (see Table 16). According to Table 16, the interaction between Hand-On-Wheel and Background-Noise messages is statistically significant with a p-value = 0.016.
Driving with One-Hand and Background -Noise (News) reaches a higher penalty score than driving with Two-Hand and Background-Noise (News).
The results also show that the interaction between the Hand-on-Wheel and the Weather is statistically significant. Driving with one hand and driving with Two-Hands and main factor Weather, level Rain has no significant difference in penalty points. The interaction between Hand-On-Wheel and Weather (Snow) is marginally statistically significant with the p-value =0.052. Driving with only One-Hand and main factor Weather, level Snow, however, resulted in a higher penalty score than driving with Two-Hands and main factor Weather, level Snow.
In the following sections the blocking factor Age is specifically analyzed.

Analysis of the blocking factor Age and modified responses
The ANOVA Table (Table 16) Table   18). The descriptive statistic of the modified responses in Table 18

Chart of Responses
The analysis of the factors can be seen in the Pareto Chart (see Figure 22). The Pareto Chart shows that the main factor Background-Noise is the only significant factor. The main factor Background-Noise has an p-value of 0.01 (see Table 19). Next, the response data set over 40 are analyzed. The Mean of this responses is 95.11 punishment points (see Table 20).  Table 20 shows that the data set Female has a total number of 94 responses. The Mean of these responses is 95.11 and the range of this data is 20 to 290 punishment points. The distribution of the responses are shown in Figure 23. The Figure 24 shows the Pareto Chart of the this analysis of this responses. Based on the Pareto Chart, the main factor combination Hand-on-Wheel and Background-Noise has the most significant input on the modified responses. But the combination Handon-Wheel and Weather is also significant. It is also shown in the ANOVA Table of the   analysis (see Table 21). The combination Hand-on-Wheel and Background-Noise has an p-value of 0.003 and is the most significant input on the result. The combination Hand-on-Wheel and Weather has the p-value of 0.044.

Discussion and Conclusion
This study aims to investigate and explore which factors could influence driving behavior.
Therefore, it considers the three main factors Hand-On-Wheel (One-Hand, Two-Hands), Background-Noise (Silent, News, Music) and Weather (Rain, Snow) and examines them.
The question of how these main factors and their interactions have statistically significant effects on driving behavior was examined in this thesis.

Assessment of driving performance
Each trip of the participants was automatically regarded by the simulator with a driving assessment which summarizes all violations committed by the participant during the ride.
The simulator assigns a score to the violations and evaluates the ride with a driving performance score. The evaluation of many offenses by the driving simulator did not seem appropriate. For example, High Speed Collision was rated by the simulator with a deduction of 5% in the Vehicle Handling Safety category (see Figure 20). The same number of points would also be deducted for Maximum Speed Limit Exceeded (see Figure   20) Another problem with the chosen evaluation technique is that it is still based on the driving assessment protocol, which the driving simulator creates after each ride. However, the simulator considers, for example, a behavior such as driving too slowly, not as a violation.
In addition, the assumption that the assessment of the simulator records all violations may not be correct. This is based on the observation of the researcher during the test with the driving simulator. For example, it could be observed that if two violations took place simultaneously, only one was recorded in the driving assessment protocol. This could be observed for example with the violations Maximum Speed Limit Exceeded and Following too close. For this reason, participants were possibly rated better than they actually drove.
Possibly the recording of the violations and the subsequent evaluation of the driving behavior with two observers is more reliable in order to determine the output factor.

Two approaches for the analysis of the responses
The analysis of the responses from the driving simulator experiment has been performed in two ways. In the first analysis (see Therefore, the study conducted a second analysis to perform with the same main factors and blocking factors as the first analysis. The difference from the first analysis is the response variable. All zero responses are not considered in the analysis. The decision for this step has two reasons. As previously mentioned, it is possible that the driving simulator does not recognize or evaluate every violation as such (see section 5.2). In addition, participants were observed by the researcher to be extremely cautious about paying more attention to driving than would has been expected under normal circumstances. In a personal conversation after the experiment, a few participants told the researcher that they saw the driving in the driving simulator as a challenge and tried to achieve a perfect driving performance score. Other participants, according to the researcher's observations, seemed to be somewhat overwhelmed by driving in the driving simulator and often drove far below the respective speed limit during the experiment. However, this described behavior of the participants does not correspond to normal driving behavior. For these reasons, these responses with zero punishment points were not taken into account in the second analysis (see section 4.3.2).

Results of the analysis: main factors
Both analyses came to the conclusion that Background-Noise was the most important factor in both approaches (see Figure 14 and Figure 19). But also the combinations of the main factors Hand-On-Wheel and Background-Noise, and Hand-On-Wheel and Weather were significant factors in the second approach. This study found in the first approach that the blocking factor Gender (p-value=0.077) is approximately significant for the responses and the second approach shows the blocking factor Age (p-value=0.048) is a significant factor for the responses.
The particular importance of Background-Noise for driving behavior is also confirmed in other tests. The acoustic distraction (see section 2.2.2) can prevent the driver from driving.

The experiments [38] [39] investigated the driving behavior of drivers who listened to
Background-Noise. The experiment of the researchers Nowosielski, Trick and Toxopeus [38] showed that playing an audio book in an uneventful driving environment improves driving performance. As soon as the situation became more complex, for example, due to traffic signs or other cars, the driving performance decreased.
The researchers Desmet and Diependaele [39] made an experiment about acoustic distracted driving and found out, the driver concentrates less on traffic-related information while he is busy listening and this increase the risk of not perceiving changes on the road directly. This was partially confirmed by experiment with the driving simulator in this thesis.
Contrary to the expectations, the level Music of the main factor Background-Noise leads to a worse driving performance than the level News. The higher amount of information in News suggests that the driving performance should be worse here. The reason could be the choice of the participants and the choice of the played news during the experiment. At first, the played news was only political contributions. This kind of news requires some interest.
Perhaps it was easy for participants who were less interested in politics or daily news to suppress the background noise. In addition, the played news was presented exclusively in English. But not all participants were native English speakers. For this reason, it may have been easier for the participant to ignore what they had heard. The played music, on the other hand, was songs that were regularly on the radio at the time of the experiment and was therefore familiar to most of the participants. For a repetition of the experiment, it could be better to choose a neutral background noise, besides music, with a topic of general interest. Also, it would have been better to play this background noise in the native language of the current driving participant.  [56] came to the conclusion that concentrating on a secondary activity already leads to a significant increase in the risk of an accident. As described above, the Background-Noise main factor is on itself a significant response factor. The decreasing concentration on the road means that dangers on the road are detected later. If the reaction time is already shortened by the reduced focus on the road, it is understandable that the driving performance decreases further if the driver is additionally manual distracted with holding a cup or something else in their hand and therefore cannot steer the car properly. Driving with only one hand reduces the driver's control over the steering wheel and thus over the vehicle, especially in sudden danger situations. This fact could be proved with the driving simulations experiment because the second approach shows that the combination of the main factors Hand-On-Wheel and Background-Noise is significant for the driving performance.

Results of the analysis: combination of main factors
In these environments, the additional distraction led to a further deterioration in driving performance. The second approach shows that the combination of the main factors Hand-On-Wheel and Weather is also significant for the driving performance. As already described, the driver is manual distracted when he is holding a cup during driving. In addition, the driver's ability to drive the vehicle becomes more difficult to maintain control if the vehicle reacts unpredictably. The weather conditions, especially black ice on the roads like in the driving simulator experiment, can lead to this. Since the vehicle's behavior is not always predictable it therefore explains why this factor is significant. The research of Nofal and Saeed [67] underline this conclusion. They found out that the seasonal weather influences the number and types of road traffic accidents over the year.

Results of the analysis: blocking factors
The first and second approaches showed that the blocking factors could presumably have a significant impact on the responses, too.
The first approach of the analysis suggests the conclusion that the blocking factor Gender  Table 12).
For the data set Female, this study found that only the main factor Weather is significant.
Other studies identified the weather as a significant factor for the driving performance.
Hoogendoorn, Tamminga, Hoogendoorn and Daamen [66] declared that the driving behavior, under adverse weather conditions for unfavorable weather conditions, have a considerable influence on the flow of traffic. The researchers Nofal and Saeed [67] showed in their paper that there are seasonal weather fluctuations influencing the number and types of road traffic accidents over the year. For example, in the summer seasons, more accidents are triggered by heavy traffic but also by intense sunlight. But, no other research with public access could be found that would underline the result of the analysis that snowy and rainy weather conditions have a higher impact on the driving performance of women than men. This has to be verified in a following experiment. The study by Laapotti and Keskinen (1998) investigated fatal control loss accidents in young male and female drivers surveyed by the Road Accident Investigation Team in Finland [68]. The results showed that in male drivers, control loss accidents were often due to speeding or alcohol influence. Typically, male drivers lost control resulting in accidents in the evening and at night. On the other hand, control loss accidents among female drivers usually occurred in slippery road conditions [68]. This supports the results of this thesis that the weather is a significant factor for the driving performance of female drivers. However, the study only examined drivers between the ages of 18 and 21. In addition, 338 men and 75 women were examined in accidents [68].
The comparison of the Means of both data sets show that the Mean of the female participants is 66.92 and 81.50 for male participants. This suggests the conclusion that the female participants drove better because they reached in average a lower number of punishment points. But the descriptive analysis of both data sets also shows that the male drivers reached 290 maximum. The female drivers with the lowest driving performance has 390 punishment points, even though this number can only be traced back to one participant (see Figure 17).
The results of the paper "Traffic Accident involvement rates by driver age and gender" of Massie, Campbell and Williams verified that men had a higher risk than women of experiencing a crash [69]. But the study also recognized that a driver's performance depends on the age of the drivers. It was recognized that older drivers had a higher fatal accident rate, while mainly young and unexperienced drivers had a high rate of involvement in crashes reported by police [69]. The mean of blocking factor groups is male Beck and Zanjani (2019) which came to the conclusion that age has a negative impact on the driving performance [73]. These results should be investigated in following studies.
Through an analysis, the significant factors on the driving performance of men and women, The mean of the responses in the data set under 40 is 79.20 (see Table 18) and the mean of the responses in the data set over 40 (see Table 20) is 95.11. This mean of the responses of the data set under 40 is 15.91 lower than over 40, which means the participants in this group had a better driving performance.
The researchers Rhodes and Pivik [71] conducted a survey in the US state of Alabama to investigate the relationship between risk perception and risky driving. They came to the conclusion that risky behavior is more likely to occur in younger drivers. Other studies also suggest this conclusion. In contrast, researcher of the paper "Vision, Aging, and Driving: The Problems of Older Drivers" [72] interviewed participants aged 22-92 about their visual impairments while driving and about performing everyday tasks. Many of the respondents in the higher ages indicated that unexpected vehicles, vehicle speed, dim displays, windshield problems, and sign reading caused them problems from time to time [72]. These problems also seemed to be related to the more frequent car accidents among older drivers.
But besides the age-related changes in visual function, the aging population also has other factors which have an impact on safe driving. A study by Allen, Beck and Zanjani (2019) investigated the driving concerns of older adults and their ability to drive [73]. They found that eighty-four percent of participants reported at least one driving problem. The most commonly reported driving problems were other drivers in general, driving at night, vision and awareness, and aggressive or reckless drivers.

Limitations of the study
In addition, this study acknowledges some limitations. These limitations could impact the reliabilities of the findings. Therefore, the limitations could be classified as the following: First, this study had a challenge to recruit in order to have enough of a homogenous sample size. Because of the length of the study, a lot of people did not want to participate or did not show up. Also, the motion sickness held some people back.
Second, the software Scenario Builder™ is limited to create experiments that identify which factor could impact the driving performance. Hence, the simulator had less sensitives to record all the drivers' violations during the experiment. Especially, the simulator had less flexibility to design the experiment scenario. For example, it was only possible to place a limited number of cars and pedestrians.
Based on the descriptive analysis, the standard deviation is larger than the mean and that can also impact the factors. Therefore, the significant factors of main and interaction factors are sensitive to the standard deviation of the responses.
Finally, the traffic policies vary from state to state, which could intern make it difficult to use our study's approach to compute the driving performance based on the punishment scoring.

Summary and Future Work
In this thesis, an attempt has been made to investigate the effects of distracted driving on In summary, the analysis results of the experiment proved that even seemingly insignificant distractions influence the driver's attention so that driving performance decreases. Even if some of the main factors did not affect driving performance alone, all of the main factors, which were investigated, analyzed in combination with another main factor had a negative effect on driving performance.
The ever-increasing number of cars and the fact that most people spend more and more time in the car is expected to increase the number of fatalities due to distractions while driving. To counteract this number, more education is needed to make drivers and passengers more aware of this problem. In the survey, which was also conducted as part of this work, 89.47% of participants, regardless of Age or Gender, stated that they listen to music or the radio and 92.98% of the participants said that they sometimes drive with only This study recommends to include and investigate more factors that could have an impact on the driving performance main factor levels and their combinations, such as speed limit (for example, above traffic speed limit, and below or equal traffic speed limit), daylight (for example, morning and night) and road types (for example, highway and side roads).
Also, for the Background-Noise level Music and News use for each participant the same audio record for both levels to control the emotional factor of the participant. Then, a study with a larger and homogenous sample number size is suggested and recommended to represent the population in a certain country or state.