Thanks @StatDave , that is definitely what I was looking for. I hate to keep asking questions on this same thread, but do you have any idea how to use Hans-Peter Piepho's mult macro to get the multiple comparison letters for those adjusted p-values? I have tried to use it but it gives me nonsensical output. Here is what I have tried: first, label the comparisons by interval so that the proc multtest does the Holm adjustment within each interval. data my_nlmeans; set est;
interval=ceil(_n_/6);
run;
proc multtest inpvalues(pr)=my_nlmeans holm;
by interval; id label;
ods output pvalues=mult_pvals;
run; Next, manually label the p-value data object and the nlmeans data object with columns for the treatments being compared, as I believe is required by the mult macro. Then run the mult macro for an individual interval. This gives me "a" for every group even though some of the adjusted p-values are below 0.05. Maybe I am using it incorrectly? data tissue_contrast_labels;
infile datalines missover;
input interval _interval response $10. _response $10.;
datalines;
1 1 berries female
1 1 berries leaftip
1 1 berries male
1 1 female leaftip
1 1 female male
1 1 leaftip male
2 2 berries female
2 2 berries leaftip
2 2 berries male
2 2 female leaftip
2 2 female male
2 2 leaftip male
3 3 berries female
3 3 berries leaftip
3 3 berries male
3 3 female leaftip
3 3 female male
3 3 leaftip male
4 4 berries female
4 4 berries leaftip
4 4 berries male
4 4 female leaftip
4 4 female male
4 4 leaftip male
;
run;
data mult_pvals; merge mult_pvals tissue_contrast_labels; run;
data my_nlmeans; merge my_nlmeans tissue_contrast_labels; run;
%mult(diffs=mult_pvals, lsmeans=my_nlmeans, trt=response, p=stepdownbonferroni, by=interval, level=1);
... View more