DATA TRANSFORMATION AND IMPLICATIONS OF USING THE KRUSKAL-WALLIS TEST IN AGROECOLOGICAL RESEARCH

In the present work, conceptual and practical aspects (hypothetical example using Lycopersicum esculentum) which must be considered in the utilization of the Kruskal-Wallis test (a nonparametric statistic test) were discussed, highlighting the possible complications when such test is arbitrarily implemented in agroecological studies. Furthermore, it was considered the equivocation that could occur when rather general assertions concerning data transformation are made, mainly about the arcsine transformation of the square root on proportion data. Careful consideration in relation to the implications of the outcomes derived from parametric and nonparametric testes should be present as an essential feature in agroecological experimental designs. As indicated in our hypothetical example, such outcomes derived from both tests might not permit the same conclusions in small sample size situations.