Beyond analysis of variance. A comparison of some multiple comparison procedures

In analysis of variance, when the null hypothesis of no difference among several treatment means is rejected, the investigator concludes that not all treatment means are equal. The logical question that arises at this point concerns which means, then, are different from which others. In many research reports, this question is never answered. In those instances where an attempt is made to pinpoint just where the differences among means occur, the procedure employed is frequently inappropriate. This article describes the most frequently used of the theoretically sound procedures that are available for making all possible comparisons among means. Four procedures are illustrated with the same experimental example.