National Academies Press: OpenBook

Design of the In-Vehicle Driving Behavior and Crash Risk Study (2011)

Chapter: Chapter 2 - Study Design

« Previous: Chapter 1 - Introduction
Page 7
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 7
Page 8
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 8
Page 9
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 9
Page 10
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 10
Page 11
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 11
Page 12
Suggested Citation:"Chapter 2 - Study Design." National Academies of Sciences, Engineering, and Medicine. 2011. Design of the In-Vehicle Driving Behavior and Crash Risk Study. Washington, DC: The National Academies Press. doi: 10.17226/14494.
×
Page 12

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

C H A P T E R 2 Study DesignThe study design elements have been defined from a variety of perspectives to try to make certain that there is sufficient statistical power to detect statistically significant effects in as many cases as possible. In this way, the data produced by the study will be able to answer as many of a comprehensive set of categorized research questions as feasible, while staying within the bounds of known project constraints such as fund- ing and time. These constraints serve to limit the project in several ways, such as the duration of the data collection and the number of data collection sites that can be established and managed. The set of research questions has guided decision-making processes throughout the design of the SHRP 2 NDS. Addi- tionally, sampling statisticians and other experts were con- sulted in the development of the sample design and contractor site-selection process. Each of these elements, together with the process for participant recruitment, is discussed below. Specific Research Questions On the basis of numerous source documents, including The 100-Car Naturalistic Driving Study, Phase II (Dingus et al. 2006), the Tri-Level Study of the Causes of Traffic Accidents Final Report (Treat et al. 1979), and Pre-Crash Scenario Typol- ogy for Crash Avoidance Research (Najm et al. 2007), as well as expertise from researchers in the transportation safety field, 392 research questions were identified for considera- tion in the NDS design. These questions address different types of crashes and other non–crash-specific safety areas of interest. They also provided the basis for making a variety of decisions regarding the number and types of DASs used to collect data during the study, the types and priority of addi- tional sensors and DAS capabilities, and the sampling and analysis trade-offs. The questions were subcategorized into the following per- spectives (the list also includes examples of study design deci- sions intended to address the questions, as applicable):7• Traffic-, roadway-, and environment-based questions. Roadway data will be collected in the Roadway Data Col- lection effort (S04). Front-mounted radar combined with the images from several synched video cameras will pro- vide traffic information. Environment-based questions can be addressed in several ways. Information will be available about time of day, ambient illuminance levels, windshield wiper status, and, for select crashes, the relevant details of the crash site will be documented by a trained crash site investigator. • Vehicle-based questions. Vehicle details, including make, model, and year, as well as all onboard safety features, will be documented for each vehicle in the study. • Driver- or driver-error-based questions. Each primary driver in the study will be characterized in terms of his or her demographic details. In addition, each primary driver will undergo a fairly extensive suite of assessments designed to characterize him or her along various driving-relevant dimensions of ability (e.g., visual perception). In addition, continuous video will be collected from four cameras. These can be reviewed for any event of interest to address error- based questions. • Passenger-based questions. A fifth camera will be aimed to capture an irrevocably blurred still image of the cabin every 10 minutes. This is done to capture basic information on the number of passengers as well as an approximation of their ages and genders for each trip. • Infotainment-system-based and nomadic-device-based questions. Information on Infotainment system status will be available from the controller area network (CAN) bus for some of the vehicle fleet. Nomadic device (e.g., handheld technology, such as a cell phone) usage can be determined using appropriate sampling schemes, though hands-free usage will be more difficult to reliably detect. • Aggressive-driving-based questions. Aggressive driving patterns can be seen in the exceedance of preset (but adjustable) levels of various sensors or combinations thereof

8as expressed in pretested algorithms (e.g., longitudinal deceleration < −0.5 g and forward time-to-collision < 3.0 s). • Vision-, attention-, and distraction-based questions. Continuous video of the participant’s face permits the scoring of glance direction and duration which can be ana- lyzed to address these areas of relevance. This can be accom- plished via manual frame-by-frame video data reduction, but machine-vision-based techniques are being developed at VTTI to enable some of the glance-related scoring to be performed in an automated fashion. • Speed- and speeding-based questions. A set of redundant sensors provides continuous vehicle speed information. In addition, onboard Global Positioning System (GPS) sensor data can be combined with GIS-based information on speed limits to permit researchers to look at speeding- related behaviors (e.g., actual speed relative to the posted speed limit) and safety-related events. • Crash-countermeasure-based questions. Crash counter- measures can be evaluated by examining actual crash-related events and superimposing different countermeasure algo- rithms to evaluate their relative effectiveness. Similarly, these same algorithms can be applied to non-event baseline epochs to determine their relative propensity for creating false alarms. • Passing-maneuver-based questions. Passing maneuvers can be detected in the data stream (i.e., via the yaw sensor) and differentiated from swerves on the basis of turn signal status and visual validation of the passing maneuver. • Multifactor/Multivariate questions. All of the above types of questions (and many more not listed) can be looked at in any combination desired by a researcher. For instance, a researcher may be interested in looking at a speeding-related countermeasure to evaluate whether it may have been pos- sible to prevent any speeding-related crash events observed. The research questions were grouped by the SHRP 2 Safety Project S02 contractor into the following high-level categories: • How does driver distraction influence crash likelihood? • How does driver fatigue affect crash likelihood? • How do aggressive driving behaviors influence crash like- lihood? • What is the influence of driver impairment (e.g., alcohol) on crash likelihood? • How do driver interactions with roadway features influ- ence the likelihood of lane departure crashes? • How do driver interactions with intersection features (con- figuration and operations) influence crash likelihood? • How do advanced driver support systems influence crash likelihood? • What variables or pre-event factors are the most effective crash surrogate measures? What explanatory factors areassociated with crashes or crash surrogates? And what ana- lytical models can be developed to predict crash or crash surrogates? The list of research questions will continue to expand as the full implications of the data are analyzed and understood. Experience with other naturalistic studies, and the expecta- tion for the SHRP 2 NDS, suggests that the data will be use- ful for addressing many questions for years after the data are collected. The specific research questions are discussed in greater detail in Appendix A of the S02 Phase 1 report (Boyle et al. forthcoming). Sample Design Plan There are two broad aspects to the NDS sample design: (1) contractor-site selection and (2) participant-vehicle sample design. The first refers to the process of determin- ing the number and location of the data-gathering sites and the contractors who will manage each; the second refers to the methods and factors used to recruit and select participants. These two aspects ultimately work toward the definition of a unified plan that specifies the quantity and characteristics of participants to be recruited at and across all site locations. From the standpoint of robust experimental design, it is desirable to ensure that there is good representation in the participant sample of the basic driver–vehicle variables of rel- evance and interest. In this study, the primary variables of interest are age, gender, and vehicle type. It has been well documented that age is a strong indicator of driving risk, with the youngest and oldest drivers having an elevated crash rate per mile driven compared with other age groups. Exposure details were carefully considered in an effort to understand the implications of driver age and gender in terms of crash rates per mile and per numbers of licensed drivers. Since the target data collection budget was already established, the following are some of the major trade-offs that were considered: • Representative versus risk-prone sample. Do we strive for a more representative sample, which enhances general- izability of the data, or do we strive for a sample we believe to be more crash-prone (i.e., one emphasizing the extremes of the age range) to observe more safety-related incidents of interest? • Overall costs of participant pay versus the ability to enhance participant attraction and retention via mean- ingful compensation amounts. It is reasonable to expect that the greater the compensation, the greater the recruit- ment uptake and retention rates would be. However, this is constrained by the project funding available for this pur- pose. It was ultimately determined that participants would

9be compensated by the nominal amount of $25 per month of participation. • Total number of sites versus number of DAS units man- aged per site. It was ultimately determined that the maxi- mum number of DAS units that could be supported was 1,950. Then, the issue becomes how many data collection sites must be established to manage those DAS units. If too many sites are established, then the costs would be prohibitive. If too few sites are established, then each may be asked to manage more DAS units than resources at a particular site may allow. • Explicitly including stratification variables in the experi- mental design versus the difficulty of filling each cell. The more variables formally included in the experimental design, the greater the risk of ending up with experimental cells with too few participants—or perhaps none at all. How- ever, if factors of interest are not formally included, then the actual sample may not include sufficient numbers for analysis, and they are likely to be unevenly spread across the other factors of interest. • Total number of primary participants versus months of data collection per participant. It is desirable to have as many participants as possible in the study, yet recruitment is expensive, and it is costly to move a DAS from one vehi- cle to another. • Total cost of data collection versus cost per data-year. Just as it is desirable to have as many participants in the study as possible, so too is it desirable to have as many years of data or data-years as possible (i.e., where a data-year is equivalent to the amount of data generated by a single par- ticipant over the course of a single year). Of course, there is a substantial cost for each study year. However, the cost per data-year tends to diminish as the study period is extended, thus making the study simultaneously more expensive yet more cost-efficient. • Site recruitment size versus a contractor’s ability to man- age the square mileage. The larger the size of a site’s recruit- ment area, the greater the probability of finding a sufficient number of participants. However, if a site is too spread out geographically, then not only will this cause the contractor difficulty in managing all the DAS and participant issues, but it will also make it more difficult for those participants at the most distant points from the installation site. Passenger cars and light trucks will be the focus of this study because these types of vehicles accounted for almost 95% of all vehicles on the road in 2007, as reported by Ward’s Automo- tive Group (2010), and 94% of all motor vehicle crashes, as reported by the National Highway Traffic Safety Administra- tion (2007). Vehicles that will be instrumented for the SHRP 2 NDS include the following: passenger cars (sedans, coupes, hatchbacks, and station wagons), pickup trucks, sport utilityvehicles (SUVs; including crossover vehicles), and minivans. It is expected that only vehicle model-years later than 2002 will be targeted for recruitment to help ensure that only mechan- ically sound vehicles with access to vehicle network data, such as speed and accelerator position, will be included in the sample. The goal is for the DAS to support installation in about half of the light-vehicle population on the road. On the basis of U.S. light-vehicle sales from model years 2000–2007, a list of the top 50 most popular vehicles were iden- tified for possible inclusion in the study. These top sellers are represented by the members of either the Alliance of Automo- bile Manufacturers (AAM) or the Association of International Automobile Manufacturers (AIAM); these organizations rep- resent a cumulative total of 89% of all U.S. light-vehicle sales for the selected time period. Relationships are being pursued with original equipment manufacturers (OEMs) belonging to AAM and/or AIAM with the objective of having CAN Param- eter IDs (PIDs) available to obtain and interpret additional data from the onboard vehicle network for high-volume models; the types of data of interest include speed, wiper usage, brake actuation, accelerator position, and turn signal usage, as well as steering data. Within the advanced technol- ogy group of participants, additional data will be collected regarding the usage of in-vehicle communication systems and advanced infotainment systems. Additionally, infor- mation about driver monitoring, feedback, and collision warning systems will be available from some participating manufacturers. Using the research questions to guide the process, the research team conducted an analysis to estimate the statis- tical power afforded by the experimental design to detect a statistically significant effect associated with the various age and gender groupings. The analysis generally indicated that the study was sufficiently powered to address the age by gender questions. This analysis was conducted on a single variable, speed variability, where previous data could be used to estimate that variable’s mean and standard deviation (both required for power estimation). Since its standard deviation was fairly high relative to its mean, this variable represents a relatively conservative estimate of the statistical power that can be expected with various analyses. Still, it must be recognized that the power associated with each analysis will be depen- dent upon the particular means and standard deviations of the measures used and the actual magnitude of the differences observed. The experimental design shown in Table 2.1 is based on the preceding investigations, and revisions were made throughout the course of the S05 study design project. Note that there are more data years than participants because some participants will participate for the full two years of data collection instead of just one. Also note that the sample design emphasizes the extremes of the age spectrum more than the middle-aged

10Gender and Primary Age Range Age Range Description One Year Two Years DAS Units Participants Data-Years M 16–17 Minor teen 72 28 100 172 200 M 18–20 Adult teen 72 28 100 172 200 M 21–25 Young adult 72 28 100 172 200 M 26–35 Adult 72 28 100 172 200 M 36–50 Middle adult 72 28 100 172 200 M 51–65 Mature adult 72 28 100 172 200 M 66–75 Younger older driver 72 28 100 172 200 M 76+ Older older driver 72 28 100 172 200 F 16–17 Minor teen 72 28 100 172 200 F 18–20 Adult teen 72 28 100 172 200 F 21–25 Young adult 72 28 100 172 200 F 26–35 Adult 72 28 100 172 200 F 36–50 Middle adult 72 28 100 172 200 F 51–65 Mature adult 72 28 100 172 200 F 66–75 Younger older driver 72 28 100 172 200 F 76+ Older older driver 72 28 100 172 200 Any Advanced vehicle technology 0 350 350 350 700 Total 1,152 798 1,950 3,102 3,900 Table 2.1. Sample Design (with Target Cell Values)groups, as the extremes are where the age-related problems tend to manifest. Data Collection Sites The objective in selecting the suite of sites for the SHRP 2 NDS was to represent to the extent feasible the wide range of driving and geographic conditions and other relevant characteristics found across the United States. Of course, the sites could not be selected independent of the site contractors: for each site selected, there had to be a suitably experienced contractor who responded to the Request for Proposals (RFP) and proposed to manage a site in a particular location. The site contractors must be technically qualified for all aspects of the study as well as have (or be able to secure) the necessary facilities to carry out the data collection plan. Note that each site selected could only represent some portion of the total environmental diversity (i.e., in terms of geographic location, predominant terrain, land use, or weather patterns, to name a few key factors). It is important for the NDS to have a broad range of environmen- tal conditions represented to understand these factors in com- bination with other driver, vehicle, and roadway factors. Potential contractors responded to two rounds of Request for Qualifications (RFQ) that addressed the qualifications of the site contractor, the characteristics of the proposed site,and the availability and quality of existing state or other local data (e.g., driver licensing, roadway inventories, driver his- tory, and crash data). Of those respondents, 11 contractors were identified as potential SHRP 2 NDS study sites. Nine of these prequalified contractors responded to an RFP issued in March 2009. Proposals specified the details regarding how proposers would go about managing one or more data col- lection sites and how the sites would contribute to the col- lective balance of the entire suite of sites in terms of the key factors such as geography, weather, state law, road types, and land usage (e.g., urban versus rural). Proposals were evalu- ated on the basis of defined contractor qualifications, includ- ing human participant research experience; well-trained, permanent, full-time technical staff; and sufficient and suit- able office, assessment, and garage space. To handle a proj- ect of this magnitude, in which a typical site will be managing up to three times as many primary participant-vehicles as the entire 100-Car Study (Dingus et al. 2006), requires substan- tial experience in working with participants and managing unanticipated circumstances. Site contractors were selected on the basis of their individual merits, as well as the collective representation they bring to the group of selected S07 sites. Figure 2.1 shows the six sites chosen to serve as the SHRP 2 NDS data collection sites. The sites are also listed below, along with the contractors who will manage each site.

11Figure 2.1. Sites for SHRP 2 Naturalistic Driving Study, indicating the number of DAS units managed per year at each site.• Erie County, N.Y.: Calspan–University of Buffalo Research Center (CUBRC); • Seattle, Wash.: Battelle Memorial Institute; • Central Pennsylvania centered on State College: Pennsyl- vania State University; • Central Indiana centered on Bloomington: Indiana Uni- versity; • Tampa Bay, Fla.: Calspan–University of Buffalo Research Center (CUBRC); and • Durham, N.C.: Westat. Participant Management This study will require the participation of more than 3,000 vol- unteers in total, spread across the study’s six data collection sites. Managing these participants from recruitment through the end of their participation will require several key activities as described below. Participant Recruitment Previous naturalistic driving studies have typically recruited volunteers through so-called traditional means, includ- ing media advertisements and posted flyers. However, a probability-based sampling approach can provide a less biased, more representative sample from which generaliza-tions can more accurately be drawn. To begin to address the questions of cost, efficiency, and measures of scope associ- ated with a probability-based sampling approach for a study of this size, a pilot study was conducted by Battelle Memo- rial Institute to identify and assess potential challenges related to effectiveness, key differences, and relative costs of using phone recruiting. Recruitment Pilot Test A work plan was developed and thoroughly tested and vali- dated by staff at Battelle with the objective of testing several major aspects of the participant recruiting and screening process. The key objectives were as follows: 1. Determine the extent to which a random selection approach can be used and how this approach compares with more traditional approaches; 2. Determine the incentives that will be required to get peo- ple to participate; and 3. Assess the participant-screening forms and participant- assessment protocols. From the outcomes of this pilot study, it is expected that an approach that combines both the call-out and traditional recruiting methods will provide a reasonable balance of ran-

12domized selection with efficient and reliable methods for obtaining the participants needed in the SHRP 2 NDS within the narrow time frame allotted. A decision was made to use a centralized call center to con- duct phone-based recruiting of volunteers to participate in the NDS. Current expectations are that, initially, the phone-based recruiting approach will be used exclusively. It may then need to be supplemented with more traditional recruiting methods centered around each site (e.g., newspaper ads and flyers placed on vehicles) to target the harder-to-fill cells in the experimen- tal design. For consistency, the call center will also field incom- ing inquiries generated by the traditional recruiting ads. By centralizing as much of the recruitment process as is practical, the recruitment process can be kept much more consistent across sites, and substantial unnecessary duplication of activ- ities across sites (such as development of recruiting protocols, materials, and data entry software) can be avoided. Consent and Privacy It is important for the informed consent process, and the privacy it guarantees to participants and their data, to be addressed in close coordination with nearly all of the other S05 tasks and across the S07 sites. To this end, none of the participants’ directly identifying information (e.g., name, address, social security number) will ever be associated with any of their data for any level of analysis. However, some participant data may inherently incorporate identifying ele- ments (e.g., face video, voice recordings, some question- naire responses, vehicle inventory, crash time and location, and some GPS information). To maintain participant privacy in light of such factors, each research project that proposes touse the SHRP 2 NDS data set must adhere to data access pro- cedures, including attaining IRB approval, that will be devel- oped as appropriate. Then, only the level of data required for the specified analyses will be made available. The main drivers of study vehicles are targeted as pri- mary participants in the study. Up to three other individu- als per primary participant, typically a participant’s family members—who can be expected to also regularly drive the study vehicle—are considered secondary drivers. Study per- sonnel will attempt to get informed consent from up to three of these secondary drivers and ask them to complete select questionnaires. However, for secondary drivers who opt out or for others who may drive the vehicle less frequently, their data that include inherently identifying information (such as facial images and GPS coordinates) will be de-identified. Efforts related to the protection of human subjects will be discussed in greater detail in Chapter 5. All human subject protocols are subject to review by the appropriate IRBs. Participant Compensation Primary drivers who volunteer to participate will be compen- sated for their participation in the SHRP 2 NDS. The data collection system will be installed in participants’ personal vehicles for either 1 or 2 years. Compensation for one year of study participation will total $300, maximum; for two years of study participation, the compensation will total $600, maxi- mum. If a participant withdraws from the study prior to the scheduled time, compensation for that participant will be pro- rated at a rate of $25 per month of participation at the time of exit. No compensation will be provided to secondary (or other) drivers of the study vehicle, whether consented or not.

Next: Chapter 3 - Data to Be Collected »
Design of the In-Vehicle Driving Behavior and Crash Risk Study Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s Strategic Highway Research Program (SHRP 2) Report S2-S05-RR-1: Design of the In-Vehicle Driving Behavior and Crash Risk Study provides a summary of the key aspects of the planning effort supporting the SHRP 2 Naturalistic Driving Study (NDS). SHRP 2 Safety Project S05: Design of the In-Vehicle Driving Behavior and Crash Risk Study (Study Design) designed the SHRP 2 NDS, which will collect data—on the order of 1 petabyte (1,000 terabytes)—on “naturalistic,” or real-world, driving behavior over a two-year period beginning in fall 2010.

The resulting data is expected to provide a wealth of information regarding driving behavior, lane departures, and intersection activities, which is anticipated to be of interest to transportation safety researchers and others for at least 20 years.

An e-book version of this report is available for purchase at Google, iTunes, and Amazon.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!