5. Multiple comparison procedures #1 Aa Aa Consider a hypothetical experiment in which the experimental units are divided into six groups, each receiving a different treatment. The six treatment groups define six populations of interest. For the experiment given, there are a total of 6 pairwise comparisons of the population means. If every pairwise comparison of the population means is tested at a significance level of o = .05, then the comparisonwise Type I error rate is .05 Assuming that each comparison test is independent of all of the others, the experimentwise Type I error rate is 05 . The experimentwise Type I error rate increases as the number of pairwise comparisons increases. Fisher's least significant difference (LSD) procedure is a multiple comparison procedure performed Suppose all pairwise comparisons in this experiment are made using Fisher's LSD procedure. If you want the maximum experimentwise Type I error rate to be .05, then the comparisonwise Type I error rate should be using the Bonferroni adjustment. For a fixed sample size, lowering the Type I error rate increases the Type II error rate