Two multivariate procedures, Automatic Interaction Detection (AID), and step-wise multiple regression, were employed to predict drinking driving traffic offences. Prediction variables includes: (1) drivers age, (2) prior traffic citations and reportable accidents, (3) criminal record, and (4) questionnaire items. It was concluded that the AID predictions were more accurate than those from the step-wise regression model. Also predictions based on questionnaire and driver-record were not improved by the addition of criminal record predictors.
Samenvatting