PREDICTING ADOLESCENT PROBLEM USE OF MARIJUANA - DEVELOPMENT AND TESTING OF A BAYESIAN MODEL

This paper reports on the development and testing of a risk assessment index for problem marijuana use designed to guide teenagers through a n extensive computer-based support system intended to help them improv e marijuana-related behaviors. Bayesian decision theory, used as the b asis of the index development process, offers the advantage of buildin g the index on subjective judgments of experts and does not require a large empirical data base. The index was found to predict an independe nt panel's ratings of teenager risk, and predict the marijuana use of 10th graders using self-reports of their profiles in the 7th grade. Im plications for future risk assessment developments are discussed.