11.1 INTRODUCTION

In many experiments in which a factorial arrangement is desired, it may not be possible to randomize completely the order of experimentation. In Chapter 8 restrictions on randomization were considered, and both randomized block and Latin square designs were discussed. There are still many practical situations in which it is not at all feasible to even randomize within a block. Under certain conditions these restrictions will lead to a split-plot design. The data in Table 11.1, which we shall use to show why such designs are quite common, summarize the effect of oven temperature T and baking time B on the life Y of an electrical component [37].

Looking only at Table 11.1, we might think of it to be a 4 3 factorial with three replications per cell and proceed to the analysis of Table 11.2. Here only the temperature has a significant effect on the life of the component at any reasonable significance level. Thus, we might explore the nature of the relationship between temperature and Y. Before embarking on such an analysis, however, we should question the...