Detecting Predictors of Rare Events: Demographic, Family, and Personal Deviance as Predictors of Stages in the Progression Toward Narcotic Addiction

Abstract

Demographic characteristics, family patterns, and personal history are the three tools we use most in attempting to explain life history outcomes, whether favorable or unfavorable. These explanatory variables usually serve us well, if somewhat monotonously, so long as the outcome is neither close to invariate nor extremely rare. Most forms of psychopathology, fortunately for society and unfortunately for solving questions of causation, tend to be rare events. When we try to predict these rare events, our variables often do not perform well at all. Finding explanatory variables means accounting for the variance, but the rarer the event we wish to explain, the less variance is there to explain. The resulting small target is hard to hit. Thus we can predict arrests, which are common, but not the particular charge for which arrested, because each is relatively rare. When the demographic, family, and personal predictors describe large segments of the population, they are certain to be poor predictors of rare events, since they select more people than will have the rare event.