Abstract
Maximum absolute errors are derived for Bartlett's chi-square approximation ior the test of the homogeneity of several variances, valid for the unequal sample sizes case. The maximum absolute error , for the relevant distributions, is given uniformly in x and depends only on two parameters—the number of populations k and the smallest sample size n of the k samples, whereas the expression on the right-hand-side of the inequality depends on k+2 parameters. For
sufficiently small, approximate but reliable critical values for the test may then be given.