ABSTRACT
Experiments in psychology frequently involve more than a single grouping variable. In the discussion of the differences between experimental and observational studies, the advantages of random allocation in the former were described: foremost among these was the balance achieved in the groups on differences between subjects existing prior to the experiment. In the 1930s Fisher introduced the technique known as the analysis of covariance as a means of reducing error variation and increasing the sensitivity of an analysis of variance for detecting mean differences. When the analysis of variance procedure is used in experiments involving random allocation of subjects to groups it is generally a completely acceptable approach to deriving more sensitive tests. One of the corner-stones of a well-designed experiment is randomization, a procedure which can only partially be replaced by methods such as analysis of covariance which purport to offer some form of statistical control.
