ABSTRACT
In previous chapters we have considered binary data and absolute and relative measures for comparing proportions. Often, the estimated difference between two proportions and the confidence limits for the difference have a very natural, direct interpretation on the original scale of proportions. The difference of proportions and its reciprocal, the number needed to treat, are absolute measures. In some situations, the relative risk and odds ratio, which are relative or scaled measures, are more satisfactory measures to interpret.
