Details
Original language | English |
---|---|
Pages (from-to) | 721-727 |
Number of pages | 7 |
Journal | Journal of applied statistics |
Volume | 33 |
Issue number | 7 |
Publication status | E-pub ahead of print - 21 Nov 2006 |
Abstract
An ordered heterogeneity (OH) test is a test for a trend that combines a non-directional heterogeneity test with the rank-order information specified under the alternative. We propose two modifications of the OH test procedure: (1) to use the mean ranks of the groups rather than the sample means to determine the observed ordering of the groups, and (2) to use the maximum correlation out of the 2k-1 - 1 possibilities under the alternative rather than the single ordering (1, 2, ⋯, k), where k is the number of independent groups. A simulation study indicates that these two changes increase the power of the ordered heterogeneity test when, as common in practice, the underlying distribution may deviate from a normal distribution and the trend pattern is a priori unknown. In contrast to the original OH test, the modified OH test can detect all possible patterns under the alternative with a relatively high power.
Keywords
- Comparing more than two groups, k-sample test, Non-parametric tests, Spearman's rank correlation, Tests for trend
ASJC Scopus subject areas
- Mathematics(all)
- Statistics and Probability
- Decision Sciences(all)
- Statistics, Probability and Uncertainty
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Journal of applied statistics, Vol. 33, No. 7, 21.11.2006, p. 721-727.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - A robust modification of the ordered-heterogeneity test
AU - Neuháuser, Markus
AU - Hothorn, Ludwig A.
PY - 2006/11/21
Y1 - 2006/11/21
N2 - An ordered heterogeneity (OH) test is a test for a trend that combines a non-directional heterogeneity test with the rank-order information specified under the alternative. We propose two modifications of the OH test procedure: (1) to use the mean ranks of the groups rather than the sample means to determine the observed ordering of the groups, and (2) to use the maximum correlation out of the 2k-1 - 1 possibilities under the alternative rather than the single ordering (1, 2, ⋯, k), where k is the number of independent groups. A simulation study indicates that these two changes increase the power of the ordered heterogeneity test when, as common in practice, the underlying distribution may deviate from a normal distribution and the trend pattern is a priori unknown. In contrast to the original OH test, the modified OH test can detect all possible patterns under the alternative with a relatively high power.
AB - An ordered heterogeneity (OH) test is a test for a trend that combines a non-directional heterogeneity test with the rank-order information specified under the alternative. We propose two modifications of the OH test procedure: (1) to use the mean ranks of the groups rather than the sample means to determine the observed ordering of the groups, and (2) to use the maximum correlation out of the 2k-1 - 1 possibilities under the alternative rather than the single ordering (1, 2, ⋯, k), where k is the number of independent groups. A simulation study indicates that these two changes increase the power of the ordered heterogeneity test when, as common in practice, the underlying distribution may deviate from a normal distribution and the trend pattern is a priori unknown. In contrast to the original OH test, the modified OH test can detect all possible patterns under the alternative with a relatively high power.
KW - Comparing more than two groups
KW - k-sample test
KW - Non-parametric tests
KW - Spearman's rank correlation
KW - Tests for trend
UR - http://www.scopus.com/inward/record.url?scp=33746146995&partnerID=8YFLogxK
U2 - 10.1080/02664760600708954
DO - 10.1080/02664760600708954
M3 - Article
AN - SCOPUS:33746146995
VL - 33
SP - 721
EP - 727
JO - Journal of applied statistics
JF - Journal of applied statistics
SN - 0266-4763
IS - 7
ER -