ESTIMATIONS OF THE INCIDENCE OF ALCOHOL (BAC) AMONG TRAFFIC VICTIMS ARE HIGHLY DEPENDENT ON THE PROCEDURES USED IN ESTIMATING THE BAC DISTRIBUTION AMONG UNTESTED CASES. THIS PAPER DEVELOPS A GENERAL INFERENTIAL METHODOLOGY FOR PRORATING UNTESTED CASES TO VARIOUS BAC LEVELS BASED ON THE MAGNITUDE OF OTHER CHARACTERISTICS THEY POSSESS. THE PROCEDURE IS DEMONSTRATED FIRST BY DICHOTOMIZING FATALLY INJURED DRIVERS INTO THOSE WHO HAD BEEN DRINKING AND THOSE WHO HAD NOT. APPLYING THE METHODOLOGY, TO DATA ON FATALLY INJURED CANADIAN DRIVERS IT IS ESTIMATED THAT 54.3% HAD BEEN DRINKING. THE PROPORTION OF DRINKING DRIVERS AMONG THE TESTED SUBSET (56.3%) IS FOUND TO BE GREATER THAN THE ESTIMATED PROPORTION OF DRINKING DRIVERS (47.9%) IN THE UNTESTED SUBSET. SUBSEQUENTLY, THE PROCEDURE IS EXTENDED BY SUBDIVIDING THE HAD BEEN DRINKING CATEGORY INTO FIVE LEVELS, COVERING THE RANGE OF BAC MEASUREMENTS. IT IS OBSERVED THAT THE LIKELIHOOD OF A BAC TEST BEING CONDUCTED IS LOWEST FOR DRIVERS WHO HAD NOT BEEN DRINKING, BUT INCREASES WITH INCREASING BAC. (Author/publisher).
Abstract