ABSTRACT
This chapter considers multiple comparision methods that control the probability of making an error in any of the tests, when all of the null hypotheses are correct. Many multiple testing procedures can be adjusted to provide multiple confidence intervals that have a guaranteed simultaneous coverage. The chapter presents multiple comparison methods in the context of the one-way Analysis of variance (ANOVA) model but the methods extend to many other situations. The name "least significant difference" comes from comparing pairs of means in a balanced ANOVA. The easiest way to adjust for multiple comparisons is to use the least significant difference method. If the absolute difference between two sample means is greater than the least significant difference, the population means are declared significantly different. The least significant difference method has traditionally been ascribed to R. A. Fisher and is often called "Fisher's least significant difference method.".
