The development and evaluation of accident predictive models
Abstract
A mathematical model that will predict the incremental change in the dependent variables (accident types) resulting from changes in the independent variables is developed. The end product is a tool for estimating the expected number and type of accidents for a given highway segment. The data segments (accidents) are separated in exclusive groups via a branching process and variance is further reduced using stepwise multiple regression. The standard error of the estimate is calculated for each model. The dependent variables are the frequency, density, and rate of 18 types of accidents among the independent variables are: district, county, highway geometry, land use, type of zone, speed limit, signal code, type of intersection, number of intersection legs, number of turn lanes, leftturn control, allred interval, average daily traffic, and outlier code. Models for nonintersectional accidents did not fit nor validate as well as models for intersectional accidents.
 Publication:

Ph.D. Thesis
 Pub Date:
 December 1980
 Bibcode:
 1980PhDT........94M
 Keywords:

 Automobile Accidents;
 Mathematical Models;
 Prediction Analysis Techniques;
 Predictions;
 Dependent Variables;
 Highways;
 Independent Variables;
 Intersections;
 Proving;
 Regression Analysis;
 Traffic Control;
 Variance (Statistics);
 Engineering (General)