Using Selection Procedures to Analyze State Traffic Fatality Rates

Abstract

SYNOPTIC ABSTRACTTwo decades (1982-2002) of state motor vehicle traffic fatality rates are analyzed using descriptive techniques, parametric indifference zone ranking and selection procedures, and nonparametric subset selection methodologies. Each of the statistical selection methods identify a state, or a group of states, asserted to be the best (or worst), or containing the best (or worst), with a prescribed probability applicable to a stated parametric configuration of the populations. Operating characteristic curves are developed for the indifference zone selection procedures. Best and worst states are defined in terms of ordered (state) location parameters of the assumed underlying probability distributions for the fatality rates. Best (worst) is characterized by the smallest (largest) location parameter. Power transformations of the rates are utilized to develop two-way models that are free of state-year interaction terms. Geographic patterns for the selected states are identified. The time stability of the geographic conclusions are examined by comparing the results of this study to those of an analysis of 1960-1976 motor vehicle traffic fatality rates appearing in a 1979 issue of Technometrics. The conclusions are also contrasted to those derived from similar analyses of alcohol-involved motor vehicle traffic fatality rates and several discordant states identified.