ABSTRACT
This chapter defines nesting and considers the implication of nesting for both the design and treatment structures. Nested models may arise in experiments restricting randomization or in certain factorial arrangements of combinations of levels. Sometimes restrictions are dictated by pragmatic considerations. When nesting occurs in the design structure, different factors may be assigned to different sized experimental units. Experiments with nesting only in the treatment structure may still have only one type of experimental unit and hence one identifiable source of error variance. Sub-sampling arises naturally when several measurements are taken on the same experimental unit. The chapter considers complete nesting with all fixed effects or some random effects to illustrate further the differences between fixed and random effects in nesting. A careful understanding of the design of an experiment, in particular where randomization is introduced, can reveal the nature of nesting in the design structure.
