A SAS Macro for Jackknifing the Results of Discriminant Analyses

Abstract

In discriminant analysis, the correlations between the variables and the discriminant functions, structure coefficients, are used in interpretation. Rules of thumb are used to set a cut-off value. It would be useful for structure coefficients to be at least statistically significantly different from zero before being used in an interpretation. Unfortunately, there are no formulas for the standard error of structure coefficients. However, the jackknife procedure can be applied to produce an estimate of the standard error. A SAS macro procedure for performing a jackknife analysis on structure coefficients is described together with issues and caveats on its use in multivariate methods.