topic Re: PROC NPAR1WAY - Stratified Hodges-Lehmann estimates in Statistical Procedures

PROC NPAR1WAY - Stratified Hodges-Lehmann estimates

DIZJ — Tue, 10 Aug 2021 13:38:24 GMT

I was able to produce a stratified Wilcoxon test, but I can't figure out if it is also possible to obtain stratified Hodges-Lehmann estimates for the location shift.

If that option is not available, I am not sure if it would make sense to compute it as a weighted sum of the location shift for each strata? I could not find any reference on that.

Program code:

proc npar1way data=input hl wilcoxon;
  CLASS treatment;
  VAR variable;
  EXACT hl;
  STRATA strat1 strat2 / wilcoxon;
run;

Thanks!

Re: PROC NPAR1WAY - Stratified Hodges-Lehmann estimates

Ksharp — Tue, 10 Aug 2021 13:43:53 GMT

Calling @StatDave

Re: PROC NPAR1WAY - Stratified Hodges-Lehmann estimates

SteveDenham — Wed, 11 Aug 2021 15:49:45 GMT

Check this reference out (ASA members can get around the paywall):

https://www.tandfonline.com/doi/abs/10.1198/tast.2010.08121?journalCode=utas20

The authors cover the HL aligned rank test, with some adaptive testing schemes, and are quite pleased with the results.

SteveDenham