|
Activity Number:
|
320
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #304739 |
|
Title:
|
Comparing Two Groups Based on Partial Orderings: Three Matched Pair Designs
|
|
Author(s):
|
Jun Cao*+ and Woollcott Smith
|
|
Companies:
|
Temple University and Temple University
|
|
Address:
|
, , ,
|
|
Keywords:
|
Partial ordering ; Sign test ; Linear extensions ; Ballot problem ; Wilcoxon test
|
|
Abstract:
|
We investigate statistical tests for the shift in distribution that are related to the classical Wilcoxon test when only a partial ordering of the observations is known. We find exact results for special case when both groups are of size n and the ranks within each group are known, but only n matched pairs between groups can be observed. Three designs are considered: the pairings are assigned at random; the pairings have the same within group rank and the pairings are assigned by the reversing the within group ranks. A surprising result is that the reverse-rank design has greater power than the paired-rank design. These designs are analogous to a team tennis tournament when player pairings are determined by their ranks within each team, usually by pairing the ranks.
|
