The One-Way Layout
The procedures of this chapter are designed for statistical analyses in which primary interest is centered on the relative locations (medians) of three or more populations. This development represents a direct generalization of the two-sample location problem (discussed in Chapter 4) to situations in which the data consist of random samples, one sample from each of populations. The basic null hypothesis of interest in that of no differences in locations (medians), under which the samples can be treated as a single (combined) sample from one population. The alternatives considered here correspond to a variety of restricted nonnull relationships between the locations (medians). We encounter two types of data for which such analyses are important. The first of these corresponds to a general setting of populations (referred to as treatments for convenience) with no additional conditions. The second deals with the setting where one of the treatments represents a control (or placebo) population, and we are interested in detecting which, if any, of ...