ABSTRACT

Two essential characteristics of a predictive model are its reliability and validity, terms that are often misunderstood and misused. Reliability refers to a model* yielding consistent results.† For a predictive model, the proverbial question is how dependable (reliable) the model is for predictive purposes. A predictive model is reliable to the extent the model is producing repeatable predictions. It is normal for an individual to vary in performance, as chance inuences are always in operation, but performance is expected to

fall within narrow limits. Thus, for a predictive model, predictions for the same individual, obtained from repeated implementations of the (reliable) model, are expected to vary closely.