Modifications of the Dorn-Stouffer-Tibbitts Method for "Testing the Significance of Comparisons in Sociological Data"

The published formulas for testing the statistical significance of the difference between the differences in two proportions, and for testing the statistical significance of other related kinds of comparisons, require modification. The present paper indicates what is unsatisfactory about the published formulas, and it presents more satisfactory modifications of them. Some actual and hypothetical data on rates of parole violation are used to explicate and illustrate the proposed modifications. If the variability between the relevant observed proportions is rather large, then the differences between the modified formulas recommended here and the earlier published formulas can be great.