Measurement Error Evaluation of Self-reported Drug Use: a Latent Class Analysis of the US National Household Survey on Drug Abuse
Latent class analysis (LCA) is a statistical tool for evaluating the error in categorical data when two or more repeated measurements of the same survey variable are available. This paper illustrates an application of LCA for evaluating the error in self-reports of drug use using data from the 1994, 1995 and 1996 implementations of the US National Household Survey on Drug Abuse. In our application, the LCA approach is used for estimating classification errors which in turn leads to identifying problems with the questionnaire and adjusting estimates of prevalence of drug use for classification error bias. Some problems in using LCA when the indicators of the use of a particular drug are embedded in a single survey questionnaire, as in the National Household Survey on Drug Abuse, are also discussed.