Bootstrap Est-ation and Testing for Variance Equality

The purpose of this study was to develop a sngle procedure for comparing population variances which could be used for distribution forms. Bootstrap methodology was used to estimate the variability of the sample variance statiatic when the population distribution WaS normal, platykurtic and leptokurtic. The data for the study were generated and analyzed using the Statistical Analysis System computing package. The bootstrap estimates of variability underestimated the theoretical variance value, and the mean square error of the estimated variance was small for both the normal and platykurtic distributions,.but large for the leptokurtic distribution. The F-ratio and chi-square test statistics were computed for comparing the variability of two popuiations using the bootstrap estimates of variance. Observed Type I error rates within two standard errors of the nominal significance level were obtaine