Abstract
A simulation study was used to determine the finite-sample properties of various estimators of a common odds ratio from multiple 2 × 2 tables. The maximum likelihood, Mantel-Haenszel, and Woolf estimators plus four variants of each were considered. The original Woolf estimator and one of its variants are dismissed on the grounds that they often cannot be calculated. In terms of minimum bias, maximum likelihood is the recommended procedure after adding .25 to each cell.