Bootstraps taxometrics: Solving the classification problem in psychopathology.

Abstract

Classification in psychopathology is a problem in applied mathematics; it answers the empirical question "Is the latent structure of these phenotypic indicator correlations taxonic (categories) or nontaxonic (dimensions, factors)?" It is not a matter of convention or preference. Two taxometric procedures, MAMBAC and MAXCOV-HITMAX, provide independent tests of the taxonic conjecture and satisfactorily accurate estimates of the taxon base rate, the latent means, and the valid and false-positive rates achievable by various cuts. The method requires no gold standard criterion, applying crude fallible diagnostic "criteria" only in the phase of discovery to identify plausible candidate indicators. Confidence in the inference to taxonic structure and numerical accuracy of latent values is provided by multiple consistency tests, hence the term coherent cut kinetics for the general approach. Further revision of diagnostic systems should be based on taxometric analysis rather than on committee decisions based on clinical impressions and nontaxometric research.