ABSTRACT
When attention shifts to more than one comparison, several issues arise. An overall test of a compound hypothesis concerning all means is highly recommended. Overall tests of difference among group means require a bit more machinery. Since there are several possible 'directions' in which group means may differ, as summarized for instance in orthogonal contrasts, it makes sense to think about the overall magnitude of differences among groups. Formal investigation of many experiments involves characterizing sources of variation in terms of mean squares for factors and factor combinations. A well-designed experiment should have sufficient resolving power to detect effects or differences which are important to the scientist. Slight imbalance in sample sizes among groups has a modest effect on power calculations. Adjustments provide a quick tool which could then be interpreted in light of other important factors, such as time and resources, when designing an experiment.
