van Elteren test (using npar1way)

[liranme]

Hello,

  I was running the following core syntax

Proc  data=XXX;

  class GROUP;

strata STRT;

var OUTCOME;

run;

For a stratified MWM test (2 groups, 2 strata).

I got the output below:

Output

My question:

We see that the (uncorrected) test-statistic Z is: (46.3 – 42.5) / 1.907524 =  1.992111

While the output Z statistic is: 2.0001

Why is the difference? (approximation/ correction? Couldn’t find any reference and would appreciate having one).

Thanks

Liran

[FreelanceReinh]

Hello @liranme and welcome to the SAS Support Communities!

Most likely this is just because the stratified rank sum statistic (46.3) is displayed rounded to only one decimal. Note that, e.g.,

(46.3153 - 42.5) / 1.907524 = 2.00013...

Put an ODS OUTPUT statement before your PROC NPAR1WAY step,

ods output WilcoxonStrataTest=want;

print the output dataset with a numeric format showing more decimals (after submitting the PROC NPAR1WAY step)

proc print data=want;
format _numeric_ best16.;
run;

and then update your calculation of the Z statistic.