“…As much as these flaws could compromise experiments, they do not cause heteroscedasticity in a data set. The major generators of heterogeneous variance for ANOVA models are the presence of zeros in the data set and the choice of treatments with previously expected discrepant answers (the so-called effect of scale), as seeds in viability extremes, i.e., 10 and 90% of germination (Bartlett, 1936;Ahrens et al, 1990;Sakia, 1992;Lúcio, Couto, Trevisan, Martins, & Lopes, 2010). The presence of zeros was the main cause of data heteroscedasticity of hard seeds of Parkia pendula and Senna macranthera, and the reason for these data to stay heteroscedastic even when transformed.…”