An alternate method for calculating an odds ratio.

Methodological problems involved in the use of the standard Woolf-Haldane analysis of epidemiological retrospective studies are examined and an alternate method of analysis is proposed. This alternate method involves a population constructed to match the cases in numerical size and to match the controls in proportion of exposures. This method allows for finer subclassification of the data and provides a meaningful summary estimate of the relative risk. The proposed method is contrasted with the Woolf-Haldane method in the analysis of the relative risk of exposure to sick pet bird versus no pet bird for adult leukemia cases versus controls. Data is from the Tri-State Leukemia Survey. Mathematical considerations involved are contained in the appendices.