ABSTRACT
The previous chapter described how to carry out analyses which adjust for mismeasurement in continuous explanatory variables, to avoid the sorts of biases described in Chapter 2. Now we turn attention to analyses which adjust for mismeasurement in binary explanatory variables, to avoid the sorts of biases described in Chapter 3. In Sections 5.1 and 5.2 we consider a binary outcome and a binary explanatory variable, with data arising from a case-control study design. In the first instance the extent of misclassification is known exactly, whereas Section 5.2 extends to the more practical case of imperfect prior information about the misclassification. Section 5.3 moves on to consider situations where there is little or no prior information about the misclassification, but several different noisy measurements of the explanatory variable are available. Section 5.4 extends further to situations where additional precisely measured explanatory variables are measured. Some brief concluding remarks appear in Section 5.5, while mathematical details are gathered in Section 5.6.
