Construct validity of heroin abuse estimators.

The construct validity of two methodologically unrelated procedures for estimating the prevalence of heroin abuse was studied using a covariance structure model-fitting approach. A factor analytic estimation procedure and a log-linear-based capture-recapture method of estimation were simultaneously applied to 24 major metropolitan areas of the nation. A test of the construct validity of these procedures for estimating the prevalence of heroin abuse was carried out in the context of structural equation models where the results of the two methods were compared directly.