Validation study methods for estimating exposure proportions and odds ratios with misclassified data.

Two methods for making adjustments to an estimate of exposure when data are possibly misclassified are discussed. One, the indirect method, is widely used but the other, the direct method, seems less well-known, despite it being a rather more obvious approach. To implement either requires some knowledge of misclassification rates and it may be possible to estimate these by a validation study. Depending on how sampling for such a study is done, one has the choice of what method to use. Formulae for the variance of estimates are derived and the precision of the methods compared. The direct approach is found to be more efficient.