topic Re: Have Dunn's Macro Output P-Value in Statistical Procedures

Have Dunn's Macro Output P-Value

lee0094 — Tue, 31 Oct 2017 21:15:27 GMT

I am using the Dunn's macro that was posted on a previous community post (https://communities.sas.com/t5/SAS-Statistical-Procedures/Dunn-s-Test/m-p/89557#M4393). I am wondering if there is a way to adapt the code so that it will output the p-value for each pairwise comparison performed? Thanks in advance.

Re: Have Dunn's Macro Output P-Value

ballardw — Wed, 01 Nov 2017 14:25:53 GMT

Probably. What have you tried.

And since the post references a ZIP file containing lots of stuff you might want to extract the bit YOU are attempting to modify instead of having us guess.

Re: Have Dunn's Macro Output P-Value

lee0094 — Wed, 01 Nov 2017 15:34:43 GMT

The exact SAS macro I am interested in is this one: https://www.cscu.cornell.edu/workshops/Nonparametric_Files/dunn%20macro.sas For each comparison, it outputs the difference in average ranks, the cutoff value for that comparison, and if that comparison is significant given a certain alpha. What I really need to know is how to use the difference in average ranks output and the cutoff output to determine the p-value.

Re: Have Dunn's Macro Output P-Value

MaryA_Marion — Sat, 13 Nov 2021 02:34:35 GMT

I have a similar question. I used the dscf option in proc npar1way and obtained the table below. Then I found the referenced dscf macro and compared outputs. The macro does not give the pvalue so I assumed that the dscf value was gaussian normal and used my calculator to find the pvalue. However looking at the macro, I think it has a studentized range distribution. For a one-way analysis of 40 observations put in 4 groups of 10 I need to input the df. That is 40-03=37?

The cutoff at alpha=.05(.05/6?) is 3.63315. The p-value computed below is .053681? Why is not not less than.05?

data temp;
qalpha=probmc("RANGE",3.74730,.,40,4);
qalpha2=1-qalpha;
proc print; run;