Pairwise multiple comparisons under violation of the independence assumption

Several multiple comparison procedures (MCPs) were compared for their rates of Type I error and for their ability to detect true pairwise differences among means when independence of observations assumption were not satisfied . Monte Carlo results showed that, if independence is not met, none of the procedures maintain controlled alpha at the chosen nominal level, neither u sing error rate per comparison or the error rate experimentwise. However, o nce the dependence of the data was corrected the Type I error rate was main tained at the same level as when the correlation was zero in all the proced ures, except for the Fisher's (1935) least significant difference procedure (LSD) and Hayter's (1986) two-stage modified LSD procedure (FH). At the sa me time, conform the correlation increased by a small amount the power rate s also, specially, when the power was examined using per-pair power.