ABSTRACT
Illuminating discussion of measurement error in case-control studies recently appeared in Chapter 9 of Keogh and Cox (2014); see also Chapter 5 of Gustafson (2004) for a more detailed treatment. There are book-length treatments about general measurement error models; see Gustafson (2004), Carroll (2006), Buonaccorsi (2010), the former in particular dealing with the case of binary exposures. Our purpose here is not to review in detail the entire field of measurement error modeling (that would take another book!), but we will give the necessary background. We note that if there is measurement error in the 190exposure, there can be many consequences, including biases in relative risk estimates, loss of statistical power, and, especially, in the case of two or more predictors measured with error, erroneous conclusions even with an infinite sample size; see Section 10.1.3 below.
