Algorithms versus models for analyzing data that contain misclassification errors.

Espeland and Hui (1987, Biometrics 43, 1001-1012) propose a methodology for analysing epidemiologic data contaminated by misclassification. They explicate their algorithm by a data set concerning cervical cancer and circumcision. We reanalyse these data using a conditional independence assumption different from theirs, and reach radically different conclusions. Espeland and Hui's methodology is a form of correlation analysis. We propose an alternative methodology based on the logic of regression analysis.