ABSTRACT
We have emphasized that distribution-free does not mean assumption free. Sometimes we assumed samples were from populations whose cumulative distribution functions differed, if at all, only in their centrality measures, e.g., in means or in medians. In other cases we assumed only that the population distributions were symmetric, in yet others that samples came from populations where any differences were, in some sense, ordered. Independence, both between observations within a sample, and between different samples, was often of major importance.
