A transformation to homogenize the variances

A type of transformation which uses a numerical relationship between means and variances per treatment and that guarantees the homogeneity of variances between treatments is presented. The Bartlett test was used to confirm the homegeneity of variances in different scales. The data used for this study are those on the biology of amaranthus, the effect of the level of leucaena in the rumen microbial population in rams and the verification of the optimum inclusion time of zeolite and the use of diets without bicarbonate of soda for broilers. The numerical transformation proposed always produces the homogeneity of variances.