EPID 301 Chapter Notes - Chapter 12: Contingency Table, Odds Ratio, Recall Bias
Document Summary
Chapter 12: differential and non-differential misclassification bias in analytical studies. Differential and non-differential misclassification bias in context. All studies that estimate associations- descriptive and analytic- are vulnerable to differential and non-differential misclassification bias. Bias comes from study-design defects (perfect classification and perfect participation = no selection bias) Bias cannot be corrected by increasing the sample size (merely result in narrower confidence intervals around an erroneous estimate, providing a false sense of precision) Bias can be prevented, but not often corrected (once bias has occurred, it cannot usually be corrected after the fact) Retrospective measurement of exposure status in case-control studies = recall bias. Questions and instruments seeking to elicit exposure history are likely more sensitive in cases than in controls, so that misclassification of exposure happens in a way that differs depending on case or control (disease) status. Magnitude and direction of differential misclassification bias.