ABSTRACT
This chapter treats the problem of estimating a proportion when the true value is subject to misclassification. After some motivating examples, Sections 2.2 - 2.4 examine the problemunder the assumption that the true values are a random sample in the sense of having independent observations. We first examine bias in the naive estimator which ignore the misclassification and then show how to correct for the misclassification using both internal and external validation data or with known misclassification probabilities. As part of this discussion, Section 2.3 delineates the difference between misclassification and reclassification probabilities. Section 2.5 extends the results to allow for sampling from a finite population, while Section 2.6 provides some brief comments on how to account for misclassification with repeated or multiple measures without validation data. Finally, Section 2.7 provides an overview of the case where there are more than two categories for the true values and Section 2.8 provides a few mathematical developments.
