Suppose you conduct a case-control study of endometrial cancer, with 120 cases and 120 unmatched controls.According to medical records (which we will assume are correct), 100 of the cases and 60 of the controls have used postmenopausal estrogens.

Suppose another researcher interviews the subjects in part (a) of the above study to determine postmenopausal estrogen use (rather than using medical records).10% of the exposed cases and 40% of the exposed controls do not recall using postmenopausal estrogen so are classified as unexposed.

(i) Reconstruct the 2x2 table to reflect this misclassification of exposure, and recompute the odds ratio for postmenopausal estrogen use vs endometrial cancer.

(ii) Why is this misclassification of exposure considered differential?

(iii) In what direction is the bias that results from this differential misclassification of exposure?

(iv) Is this the usual direction of bias from differential misclassification?