Continuous variable transformations

Some of the continuous variables were combined into groups or categories to better describe these factors in an injury severity model. Some of the issues that will be discussed include scaling, valid values (range), and justification for use of continuous distributions for some variables that could have been discrete. In addition, this section briefly outlines the reasoning for the categories selected for the models. Among the main methods used were:

exploratory regression analysis, categorical data analysis, results from previous studies, design guidelines and standards. These methods were not used in an absolute fashion, rather a scientific process of inquiry through a series of steps led to an informed decision. Driver, road and environment-related continuous factors were transformed from continuous to categorical in order to enhance their interpretative power in the injury severity models.

The driver age variable has been briefly discussed previously. The choice of driver age groups was primarily based on the previous studies using Florida crash and driver data, such as the one by Abdel-Aty et al. (1998). The cutoff values for drivers as very young (15-19 years old), young (20-24 years old), middle (25-64 years old), old (65-79 years old), and very old (80-98 years). The use of 99 years was avoided due to the negligible frequency (less than 1%) and possible data error for unknown values.

The median width (median size) of the multilane roads should be as wide as practical to decrease the risk of head-on crashes and headlight glare at night. The Florida Greenbook standard calls for minimums as indicated in Table 5-27 below. Based on these standards and after initial tests, the median widths equal of greater than 40 ft were tested against those less than 15 ft, between 15 and 19.5 ft; and between 19.5 and 40 ft.

Table 5-27 Minimum Median Width for multilane facilities (Florida Greenbook, 2005)

Rural Highways

Design Speed (mph) Minimum Width (ft)

55 and Over 40

Under 55 22

Urban Streets

Design Speed (mph) Minimum Width (ft)

50 19.5 45 or Less 15.5

40 or Less ** 10

** Paved medians used for two-way turn lanes or painted medians

The Skid Resistance number discussed previously was analyzed using different cutoff values based on the FDOT guidelines. Recall from section 5.2.2.1 that skid resistance is expressed as the static friction coefficient multiplied by 100. Many highway agencies use the friction measurements for the purposes of rehabilitation, reconstruction and resurfacing of pavements. There are monitoring programs in place to take action at certain predetermined

intervention levels. The FDOT has a skid hazard elimination program has a systematic skid test program which covers about 25-35 percent of the Interstate and Primary Systems per year, along with new pavements. Also, District Safety Engineers use a report of wet weather crashes and determine which sections of highway with 25 percent or more wet weather crashes need skid tests. Depending on the test results, the FDOT guidelines are used to determine whether the skid hazard warrants an improvement project, there is need for further review (for new pavements up to 18 months) or if the skid test indicates the pavement friction is acceptable. The FDOT skid resistance guidelines in the Skid Hazard Reporting System Manual are shown in Table 5-28 below. (FDOT, 2006)

Table 5-28 FDOT Friction Number guidelines (State Safety Office, 2006)

All highway sections surfaces

1 2 3

QUESTIONABLE REVIEW DESIRED Posted Speed Limit

(MPH)

FN40 FN40 FN40

Less than or equal to 45 25 26-28 30

Greater than 45 27 28-30 35

The values for the groups considered were analyzed and tested in preliminary regression models. Part of this analysis is shown in Figure 5-2 and Figure 5-3 below, which confirms the general increase in severe injuries as the skid number increases. Examining the rural area distribution, we find that there is an unexpected increase in the proportion of severe injuries after skid number 35. Meanwhile the urban area distribution behaves a lot closer to expectations, lowering the severe crash proportions after skid number 35. Additional implications of these findings are discussed in the final analysis section.

All involvements

Figure 5-2 – Distribution of severe injuries by skid resistance and land use for all involvements

Wet pavement involvements

Figure 5-3 – Distribution of severe injuries by skid resistance and land use for wet pavement involvements

The analysis of Figure 5-2 and Figure 5-3 above confirms that the general increase in severe injuries as the skid number increases for both the wet pavement crashes and the total involvements. The wet pavement severe injury to driver involvement ratio is generally lower than those for the total crash involvements. This comparison is necessary to compare the skid resistance to the type of crashes that it aims to reduce. The relationship seems to hold for both wet pavement and all involvements. The implications of the sudden increase after skid resistance number 44 are discussed later on.

The surface width (nSURWIDTH) variable was used in an interaction with number of lanes to derive a lane width variable. According to the RCI Field Handbook the surface width is measured across the traveled way, not including the shoulders (FDOT, 2008). The lane width was computed by dividing the surface width by half the number of lanes. The lane widths, as recommended by the Florida greenbook are 12 feet for major arterials and 11 feet for minor arterials. The minimum width in the standard is 10 feet; however there is a significant number (8.59%) of driver crash involvements in roads with lanes less than 10 feet wide, as shown in Table 5-29 below. This is a particular concern for the sections of road that are not complying with the current standards and their effect on the safety performance of important arterial corridors in the state.

Table 5-29 Driver crash involvements in multilane high-speed roads by lane width group Lane width group Frequency Percent

Lane width < 10 ft 9323 8.59

10 ft ≤ Lane width < 11 ft 6252 5.76

11 ft ≤ Lane width ≤ 12 ft 80889 74.56

Lane width > 12 ft 12030 11.09

Total 108494 100

The Florida greenbook recommends shoulders 10 feet wide in all roads. A minimum of 6 feet is required for roads with open drainage, while 8 feet is required for roads with heavy traffic volumes or a significant volume of truck traffic. The cutoff values for the shoulder width groups reflect this policy. An overwhelming majority (89.52%) of the driver crash involvements occurred on roads with shoulder less than 6 feet wide, as shown in Table 5-30 below. Very similar proportions were found for the severe injuries. Another concern is that roads with larger shoulders have a little higher proportion of driver involvements. The wider road space has been associated with higher operating speeds, which in turn result in higher severe injury counts.

Table 5-30 Driver crash involvements in multilane high-speed roads by shoulder width group Shoulder width groups Frequency Percent

Shoulder width < 6 ft 97120 89.52

6 ft ≤ Shoulder width < 8 ft 3548 3.27

8 ft ≤ Shoulder width < 10 ft 3128 2.88

Shoulder width ≥ 10 ft 4698 4.33

The time variable was reduced to a binary variable denoting what is generally considered day and night hours after evaluation of exploratory analysis. There are some perceived negative effects of the involvements at night in severe injuries, as shown in Figure 5-4 below. This variable did not show a strong correlation in the exploratory analysis or the preliminary analysis.

The increased importance of the road characteristics versus the environmental variables might be triggered by underlying correlations between weather characteristics and road characteristics. For example, road lighting is correlated to visibility at night and if this variable is significant denotes in part the effect of night crashes in the overall safety performance of multilane high speed arterials.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Non-severe Severe

Driver injury severity

Proportion of driver involvements

6 AM- 6 PM 6 PM- 6 AM

Figure 5-4 – Distribution of severe injuries by time of day