Several procedures of multiple-comparison have been used to explain the response of treatments applied in experimental units of agronomic experiments. Among these are the mean comparison tests, sometimes appropriate for some kinds of treatments and completely inappropriate for others. The use and misuse of these tests were surveyed and assessed in the articles published in the journal Horticultura Brasileira from 1,983 to 2,000. Alternative procedures are suggested for situations in which mean comparison tests are inappropriate. We observed that in 294 articles some kind of test was used for mean comparison. In 65,6% of these articles multiple comparison tests (as Tukey, Duncan and LSD) were appropriately used for the type of involved data. In another 22,8% tests were used in a partially correct form. For the 11,6% remaining articles multiple comparison tests were inappropriately used for the type of treatments used in the research.
Qualitative treatments; quantitative treatments; contrasts; multiple comparison tests; regression techniques
