Ramona Kim is a California Highway Patrol (CHP) officer who works in the city of San Diego. Having lost her own uncle in a car accident, she is particularly interested in educating local drivers about driver safety. After discussing this idea with her commanding officer, she learns that since 2005 the CHP headquarters receives traffic-related collision information from local and state agencies and makes them publicly available on its website. Her commanding officer recommends that she focus on car accidents that result in deaths and severe injuries and encourages Officer Kim to share her findings with her colleagues as well as make presentations at local community meetings to raise awareness about car accidents among local drivers.

Many factors contribute to car accidents. Bad weather, reckless driving, distracted driving, lack of visibility at night, and failure to observe traffic laws can all lead to car accidents. In 2013, there were 3,000 deaths caused by motor-vehicle-related accidents in California. That is almost 10 motor-vehicle-related deaths per day. Conventional wisdom tells us that to stay safe, one should be careful in severe weather or while driving at night, refrain from driving under the influence, and always observe traffic laws. Beyond these conventional safety rules, it is critical to gain a better understanding of the different circumstances under which car accidents result in fatalities and severe injuries. 

Relevant San Diego traffic-accident data are extracted from the Statewide Integrated Traffic Records System database from January 1, 2013, through February 28, 2013. It is the rainy season in California during this time period; thus, drivers often have to travel in heavy rain and other bad weather conditions. Table 11.13 shows a portion of the data from the 785 accidents 

TABLE 11.13 San Diego Traffic-Accident Data, January–February 2013 

that occurred in the city of San Diego over this time period. The five variables of interest for the analysis include the day of the week (WEEKEND =1 if weekend, 0 otherwise), crash severity (CRASHSEV =1 if severe injury or fatality, 0 otherwise), whether or not there was inclement weather (WEATHER = 1 if inclement, 0 otherwise), whether or not the accident occurred on a highway (HIGHWAY = 1 if highway, 0 otherwise), and whether or not there was daylight (LIGHTING = 1 if daylight, 0 otherwise). For example, the first observation represents an accident that occurred on a weekday that resulted in a severe injury or fatality. At the time of the accident, the weather was not inclement, but there was no daylight. And finally, the accident did not occur on a highway. 

Agglomerative clustering analysis is performed using the matching coefficients as the distance measure and average linkage for the clustering method. Because the values of 0 and 1 (e.g., for the WEEKEND variable, 0 = weekdays, and 1 = weekends) are equally important in this analysis, the matching coefficients distance measure is appropriate. Figure 11.17 shows the resulting dendrogram plot.

FIGURE 11.17 Dendrogram for traffic-accident data 

A careful inspection of the dendrogram suggests that the 785 observations should be grouped into four clusters, as shown in Figure 11.17. Each cluster forms a meaningful group of observations that are distinctive from one another. The cluster membership is assigned to each observation. In order to gain a better understanding of these clusters, the average value of each variable is calculated for each cluster. These averages are shown in Table 11.14. The last column shows the number of cases in each cluster. 

TABLE 11.14 Clustering Results for Traffic Accidents in San Diego 

Based on these results, the following observations can be made. 

1. All of the accidents that resulted in fatalities or severe injuries are grouped into Cluster 1 (53 cases) and Cluster 2 (8 cases). 

2. An overwhelming majority of fatal and severe-injury cases in Cluster 1 occurred when the weather conditions were clear. 

3. Surprisingly, most fatal and severe-injury cases occurred on a weekday; only 32.1% of cases in Cluster 1 and 25% of cases in Cluster 2 occurred over the weekend. A possible explanation is that there are more weekdays than weekends in a week and, therefore, weekdays would naturally have a larger portion of the total fatal or severe-injury cases. 

4. Cluster 2 consists of eight cases that involved fatalities or severe injuries. All eight cases happened during bad weather conditions and after dark, and the majority of them occurred on the highway. 

5. The vast majority of the accidents are grouped into Cluster 3. These accidents tended to happen during the weekdays on nonhighway roads when the weather condition was not inclement. None of them resulted in fatalities or severe injuries. 

6. All the accidents in Cluster 4 happened on the highway. However, none of them resulted in fatality or severe injuries. Comparing Cluster 4 to Cluster 2, where the majority of the accidents also happened on the highway, but all cases resulted in fatality or severe injuries, we notice that the key differences between the two clusters are that all the accidents in Cluster 2 happened on weekends during the day when the weather was not inclement. While none of the variables may be a strong indicator for fatal or severe-injury cases on its own, a combination of traveling on the highway after dark on weekdays during bad weather seems to increase the chance of an accident resulting in fatality or severe injuries. 

The key finding from the cluster analysis is that many fatal and severe-injury accidents occur on weekdays when the weather condition is clear. Perhaps, when the weather condition is favorable, drivers tend to speed, causing severe injuries or fatalities in an accident. During the weekdays, drivers might also be in a rush to get home or to get to work, adding to the severity of a car accident. This study can be used to design an educational campaign to raise awareness among local drivers in San Diego.