• Federal Highway Administration >
• Publications >
• Research Publications >
• 98133 >
• Ch04_04.Cfm >
• Accident Models for Two-Lane Rural Roads: Segment and Intersections

Accident Models for Two-Lane Rural Roads: Segment and Intersections

4. Analysis

Summary

A wide variety of variables have been introduced in this chapter to facilitate the modeling in the next.

The summary univariate statistics for these variables (Tables 2 through 7) indicate that most of them show a good range of values that will provide variation for the modeling. Exceptions are: lighting along the segments (the vast majority have none), right turn/acceleration lanes on the minor legs of intersections (most have none), and intersection angle deviation from 90° on Washington State intersections. Most Washington intersection angles are 90° , perhaps in part because photolog estimates had to be used in Washington State and are much cruder than those obtained from Minnesota plans.

Bivariate statistics indicate that commercial traffic on two-lane segments correlates negatively with accidents while surface width and lane width have unexpected effects in Washington State. Traffic is the dominant variable for intersections, but the existence of a right turn lane on the major road correlates positively with accidents on three-legged intersections.

Bivariate relationships between accident variables and highway variables should be interpreted with caution: they may indicate that the highway variable correlates with a another influential highway variable. Modeling with several variables simultaneously may permit greater insight into the relative effects of different highway variables.