Abstract
Several tests (e.g., Bartlett's test, Cochran's test, Hartley's test) are available in the literature for testing (Ho) the equality of several variances of normal populations.
These tests demonstrate only the statistical significance of variances. Using the cube -root transformation of Wilson and Hilferty (1931), Rao and Harikrishna (1997) developed a graphical procedure for testing Ho. In this paper, using the log-transformation of
Bartlett and Kendall (1946), a new graphical procedure is proposed for testing Ho
These graphical procedures demonstrate the statistical as well as engineering significance of variances. The two graphical procedures are compared in the case of equal sample sizes and observed that the new procedure is better than Rao and Harikrishna's procedure when the sample size is greater than ten in each sample