S. Yitnosumarto, M.E. O'Neill
Levene's tests of homogeneity of treatment variances in completely randomised and randomised complete block experiments are examined. These tests are essentially standard analysis of variance F‐tests performed on functions of the absolute values of residuals. It is found that in order to achieve (i) equality of component mean squares under the null hypothesis, and (ii) nominal significance levels, the various standard degrees of freedom need to be modified. Copyright © 1986, Wiley Blackwell. All rights reserved
Facultas Pertanian, Universitas Brawijaya, Malang, Indonesia; Biometry Section, University of Sydney, Australia
