Approximate Confidence Limits for the Relative Risk

Abstract

The relative risk or odds ratio has been proposed as one alternative to the difference in proportions in measuring the degree of association in two by two tables. This paper proposes two new methods for finding confidence limits on the relative risk, one appropriate for large numbers, and a second for small numbers. The first is related to the usual chi-square test and the second to an approximation to the Fisher-Irwin test. The methods are compared to those previously proposed by Cornfield (1956), Cox (1958), and Woolf (1954).