ABSTRACT
The design principle of blocking, used to control for known or expected heterogeneity (variability) among experimental units, was introduced in Section 3.1.3. The basic approach is to group, or block, together sets of experimental units expected to have similar responses in the absence of different treatments, and to separate those units expected to have different responses. Blocking is frequently used in designed experiments to account for heterogeneity due to the location or timing of measurements. For example, in a glasshouse experiment (in the northern hemisphere), we might expect plots closer to the south wall of a compartment to be warmer than plots closer to the north wall, and so, we group our experimental units based on their distance from the south wall. Similarly, if samples have to be processed, but only half can be done in the morning and the remainder done in the afternoon, then the morning and afternoon sessions might be used as two blocks to guard against systematic differences caused by any change in the background conditions, or a change of the experimenter. Blocking is also widely used in observational studies. For example, if an ecological study makes observations of the species present on pairs of elds (e.g. one growing wheat, another growing oilseed rape) on several farms, then the farms can be included as a blocking structure to account for the many expected differences (caused by a combination of location and management practices) between farms. Full specication of an experiment therefore requires knowledge of both the blocking and treatments present. In developing ideas for designs with blocking, we consider a single set of treatments, which may mean either imposed treatments in a designed experiment or groups in an observational study, as in the previous chapters.
