* Input the data;
data eg11_4;
 input group $ cycles;
 cards;
new 7651
new 8337
new 6989
new 9568
old 9477
old 13581
;
run;
/*
Proc npar1way is a SAS procedure to run various distribution-free analyses.
Here we use it to do the Wilcoxon rank sum test. Note that in contrast to
what I have in the class notes, SAS always (1) takes population 1 to be the population
corresponding to the smaller sample, and (2) always takes the test statistic to be
the sum of the ranks corresponding to population 1. So here the value w*=10 (sum of ranks
for pop 1) and the p-value is P(W>w*).

The hl and alpha=.05 options tell SAS to compute the point estimate and large-sample 95% 
confidence interval.
*/
proc npar1way data=eg11_4 wilcoxon hl alpha=.05;
 class group;
 var cycles;
 exact wilcoxon;
run;
/*
To get the exact confidence interval as well, we run it again.
Note that the confidence level was found from the table: .134=1-.866 
(SAS apparently can't figure it out by itself!). That the exact 86.6% 
interval is the same as the large-sample 95% interval is a coincidence.
*/
proc npar1way data=eg11_4 wilcoxon hl alpha=.134;
 class group;
 var cycles;
 exact wilcoxon hl / alpha=.134;
run;