Imagine that all of the cases in a case–control study had their blood pressure measured by a single doctor at the local hospital but, for practical reasons, the controls had their blood pressure...

Imagine that all of the cases in a case–control study had their blood pressure measured by a single doctor at the local hospital but, for practical reasons, the controls had their blood pressure measured by their local doctor. In this situation it is likely that there would be less random error in the blood pressure readings for cases that came from a single doctor than in those for controls that came from a number of different doctors

(a) Recalculate the results of the hypothetical case–control study shown in Table 7.4 assuming that the measurement of exposure among cases was perfect but 20% of exposed controls were randomly misclassified as unexposed, and vice versa. (b) Is this misclassification differential or non-differential and why? (c) What effect has it had on the odds ratio and why? (d) What would the effect have been if we had misclassified cases instead of controls?

Table 7.4