Jackknife for Variance Analysis of Multifactor Experiments.

Abstract : A method is presented for analyzing effects of multi-factor experimental treatments on the variance of a dependent variable. The method is based on the statistical jackknife. It enables the analyst to enhance the power of an analysis by using the degrees of freedom associated with the random factor (e.g., subjects, in a behavioral experiment) of a multifactor design. The method is suitable for investigating hypotheses about trends of variances. A computer program is appended which calculates the jackknife variance estimates and other useful statistics. (Author)