M Wong1, N E Day. 1. Department of Mathematics, Hong Kong University of Science and Technology, PRC.
Abstract
BACKGROUND: In epidemiological studies, estimation of disease exposure associations will be biased if the exposure is measured with error. In previous papers, we considered the validity of an estimator of the correction factor under a variety of assumptions. In particular, in both univariate and bivariate cases, we considered the error in estimating the correction factor induced by incorrect assumptions on the independent errors of repeated measures, or of different types of measures. METHOD: We concentrate our discussion in this paper on the optimal design of the validation study based on the asymptotic variance of the estimate of the correction factor. Only the univariate situation is considered. We also present an example to illustrate the importance of suitable design. RESULTS: The value of a good biomarker is demonstrated again.
BACKGROUND: In epidemiological studies, estimation of disease exposure associations will be biased if the exposure is measured with error. In previous papers, we considered the validity of an estimator of the correction factor under a variety of assumptions. In particular, in both univariate and bivariate cases, we considered the error in estimating the correction factor induced by incorrect assumptions on the independent errors of repeated measures, or of different types of measures. METHOD: We concentrate our discussion in this paper on the optimal design of the validation study based on the asymptotic variance of the estimate of the correction factor. Only the univariate situation is considered. We also present an example to illustrate the importance of suitable design. RESULTS: The value of a good biomarker is demonstrated again.