Abstract
In highway safety studies, the accident rates are derived from the number of accidents divided by the section length that varies from a fraction of a mile to several miles. The section length affects the rates and the rates influence the accident prediction models. The rates from short sections may give "misleading" statistics and correlation to roadway and traffic conditions. This study analyzes the influence of section lengths on accident rates and how that influences the geometric variables used in accident prediction. The analyses are for rural interstate highways and rural two-lane highways. To get reliable accident prediction models the rates should be computed based from sections 0.5 mi (0.8 km) or longer. (A)