Which of the following statistics about teenage drivers is false?

Throughout the United States, motor-vehicle collisions continue to be a concern across multiple disciplines due to the alarming number of lives lost each year on our nation's roadways. These alarming rates are disturbingly high for America's youth, ages 13–25, as motor-vehicle collisions continue to be the leading cause of injury and death for this age group. The CDC cites riding and driving in a car to be the #1 threat to teen safety (2016). Recent crash data revealed that 1908 young drivers (ages 15–20) lost their lives in 2016, showing almost no improvement from the previous year (NHTSA, 2016). On a per-mile basis, teens drive fewer miles per year than other age groups, yet they experience the highest overall crash rates (NHTSA, 2016). The total population for this age group has decreased from 2007 to 2016 by 3.4%; despite this, there has been an increase in the number of young, licensed drivers (NHTSA, 2016). As of 2016, young drivers account for 5.4% of all licensed drivers in the United States– a 2.1% increase from 2015; but this age group represents 9% of all drivers involved in fatal crashes (NHTSA, 2016). Research into the causes of teen crashes has built a formidable case identifying driver inexperience along with several other variables such as teen passengers, nighttime driving, and lack of seat belt use as the primary contributors to an increased crash risk (Williams, 2003). It has been proposed that an understanding of crash potentials (an alternative term of the word “risk”) influences driving behaviors because younger teen drivers are more likely to underestimate crash potentials and, thus, more likely to engage in perilous driving behaviors (Brown and Groeger, 1988, Jonah and Dawson, 1987, Rhodes and Pivik, 2011). There is a need for a robust study in understanding the key contributing factors and patterns of these factors that trigger teen driver involved crashes.

Conventional statistical modeling techniques pay little attention to data visualization. Correspondence analysis (CA) can help mitigate this gap by analyzing the hypothesis testing in order to identify patterns of association in the data. CA can easily accommodate various scales of dataset sizes. The target of CA is to minimize the loss of information so that the maximum amount of information is retained. There is a new sturdy–robust–resistant variant of CA called Taxicab Correspondence Analysis (TCA). It can smoothly handle an abundance of data and produce satisfactory and meaningful results in the presence of outliers. Application of CA has been widely used in survey analysis. This paper focuses on a significantly large dataset (information regarding approximately 88,000 respondents) of teen survey data collected in Texas to investigate the understanding of crash potentials among teen drivers. The application of TCA on this dataset was deemed appropriate due to the method's suitability in tackling the research problem related to this large data set.

Earlier work and research context

Research on young drivers has yielded a vast amount of information regarding the contributing causes to increased crash risk and the time period in which teens are most at risk for a crash. Newly licensed drivers are at the highest risk for crashes within the first six months of obtaining their license (Mayhew et al., 2003, McCartt et al., 2003a). This risk decreases as they gain more driving experience, but conversely, research shows that a teen's transition from being newly licensed to

Survey design

The survey instrument used in this study was developed by the Teens in the Driver Seat (TDS) staff to obtain traffic safety knowledge and patterns of driving behaviors prior to intervention, as well as facilitate longitudinal analyses over time. The survey instrument has also been approved by the Texas A&M Institutional Review Board. The survey was four pages long and consisted of 12 main questions (see Fig. 1). Part One (questions 1–5) of the survey obtained demographic information including

Descriptive statistics

The final dataset contains 88,065 respondent data with 23 questions answered completely or in part. The survey contains responses for 28 questions, or parts of the questions (question 6 has 5 parts, and question 12 has 13 parts). Responses regarding several questions (question 1, 2, 4, and 8) were not included in the final analysis due to redundancy or a larger number of missing values. For example, grade level is highly correlated with licensing status, so it was unnecessary to include both

Theory

In a series of papers, Choulakian explained the extended theory of TCA (Choulakian, 2006a, Choulakian, 2006b, Choulakian, 2013). This section of the paper provides a brief overview of the basic concepts of TCA. Correspondence analysis (CA) is based on Euclidean distance, whereas Taxicab correspondence analysis (TCA) is based on the Manhattan, City Block or Taxicab distance. Let X = (x1, x2, …. ., xn) and Y = (y1, y2, …. ., yn) and a vector v = (v1, v2, …. ., vn) to evaluate these distances:Euclidean

Results and discussions

Correspondence analysis has recently been gaining popularity among archeologists and it is often applied to archeological abundance data. Sometimes data sets are sparse, where the degree of sparsity is defined as the percentage of zero abundances. For sparse data sets, three types of potential outliers may be identified: rare observations, zero-block structure, and relatively high valued cells. These multivariate statistical methods summarize the knowledge extraction from a complex dataset set

Conclusions

The teen survey data collected from Texas showed interesting trends and differences between respondents of the four categories of license statuses. This study found that males with provisional or unrestricted licenses were among the highest risk group, which is consistent with previous research findings. The transition to an unrestricted license has been shown to increase crash risk and, as the data reflects, these drivers reported engaging in higher risk driving behaviors more frequently than

Acknowledgement

The authors are grateful to an anonymous reviewer for providing useful comments and suggestions.

Subasish Das is an Associate Transportation Researcher at Texas A&M Transportation Institute (TTI). He received his Ph.D. in Civil Engineering from University of Louisiana at Lafayette in 2015 and holds a Master of Science in Civil Engineering from the same university in 2012. He has more than 10 years of national and international experience associated with transportation safety engineering research projects. His primary fields of research interest are roadway safety, roadway design, and

Cited by (16)

Recommended articles (6)