Abstract

In the analysis of variance, the comparison of means is essential when the calculated F is significant at 0.05 or 0.01 probability level and there are more than two treatments, because the significant F only rejects the null hypothesis, according to which the treatments or samples do not differ statistically. This study was aimed at to evaluate the similarities and differences between the classifications of means of Tukey, SNK, Scott-Knott and Duncan tests, as well as to demonstrate the performance of the software Assistat in the analysis of experimental data of the agricultural research. Data of agricultural experiments were analyzed using the models of the analysis of variance (ANOVA), as completely randomized and randomized block experiments. It was concluded that the Tukey test provides more-detailed results in comparison to the tests of Duncan, Scott-Knott and SNK, but not very different, and it is the most used test. The tests of Duncan, Scott-Knott and SNK tend to show similar results, except for the fact that, in the Scott-Knott test, no mean can belong to more than one group. The tests of Duncan and SNK, for being similar, except for the utilized distribution, almost always show the same results.  

Key words: Assistat, Tukey test, Duncan test, Scott-Knott test, SNK test.