Toward a second generation of statistical prediction rules in psychodiagnosis and personality assessment

摘要

Limitations of present-day statistical prediction rules are described. When rules for making diagnoses and describing personality and psychopathology are constructed, there is frequently no attempt to identify the best input information that is available. Furthermore, we do not know how accurate these statistical rules are because they have almost always been used to predict criterion scores that are of marginal importance and because they have been compared to clinical judges who have been given little information. Recommendations are made for (a) changing the way we build statistical rules and (b) changing the way we evaluate the validity of the rules.