Estimation of measurement bias in self-reports of drug use with applications to the national household survey on drug abuse