U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-RD-98-133
Date: October 1998
Accident Models for Two-Lane Rural Roads: Segment and Intersections
6. Validation and Further Analysis
This chapter is devoted to miscellaneous analytical tasks relevant to possible uses of the models:
Validation tests are performed to measure the predictive efficacy of the leading models.
The Minnesota models are tested against Minnesota data from a later time period (1990-1993) on the
same segments and intersections. They are also tested against Washington data, and the Washington segment model and the combined segment model are tested on Minnesota data from 1985-89 and 1990-93.
The relative explanatory value of different groups of variables in the final models (Tables 27 and 35) is assessed by means of the Log-Likelihood R-squared introduced in Chapter 5.
Scaled residuals (observed accident counts minus predicted mean accident counts divided by estimated standard error) are compared graphically with leading variables to check for
systematic trends that might contradict the assumed model form or suggest model refinements.
Explanatory Value of Final Models
Cumulative Scaled Residuals