The objective was to verify the efficiency and uniformity of responses from different tests to compare means and the rate of occurrence of Type I error. Four different tests for multiple comparison of means were applied to 200 experiments and 10 variables created with random values, to assess the number of levels; agreement with the F test and rate of occurrence of type I error per variable and per experiment. The tests differed in their results for all variables. The greatest divergences in results were found when using the LSD test. For the level of significance, the greatest agreement was found by the Tukey test.
type I error; Tukey; Duncan; Dunnet; LSD
