You can use the PROC MULTTEST to address the multiple testing problem. PROC MULTTEST can adjust the raw p-values from multiple tests by using any of several adjustment methods. Note that the Bonferroni method is quite conservative so you might want to consider other adjustment methods. PROC MULTTEST can perform certain tests on the raw data to generate the p-values to adjust. For tests that are not available in PROC MULTTEST, you can provide an input data set of p-values for it to adjust. For continuous response data, PROC MULTTEST can perform t-tests. The tests that PROC MULTTEST can perform on discrete response data are the Cochran-Armitage trend test, the Freeman-Tukey double arcsine test, the Peto (Logrank) test, and the Fisher exact test.

The following is an example of using PROC MULTTEST to adjust a set of raw p-values in an input data set. Three adjustment methods — Holm, Hochberg, and Benjamini (FDR) — are used. The variable holding the raw p-values and the name of the data set containing that variable are specified in the INPVALUES= option. You can also create a data set containing the raw and the adjusted p-values by using the OUT= option in the PROC MULTTEST statement.

      data a;
         input Test $ pval;
         datalines;
      test1 .09108
      test2 .69122
      test3 .00177
      test4 .57181
      test5 .03121
      test6 .01413
      ;
      
      proc multtest inpvalues(pval)=a holm hoc fdr;
         run;

Operating System and Release Information