Pragmatics of Pooling in ANOVA Tables
作者:
W.G. S. Hines,
期刊:
The American Statistician
(Taylor Available online 1996)
卷期:
Volume 50,
issue 2
页码: 127-139
ISSN:0003-1305
年代: 1996
DOI:10.1080/00031305.1996.10474360
出版商: Taylor & Francis Group
关键词: Conditional pooling;Increased error degrees of freedom;Nonsignificant interactions;Taguchi methodology
数据来源: Taylor
摘要:
Pragmatically, is “pooling” worth doing in analysis of variance (ANOVA) tables, statistical niceties aside? The conditions found for pooling to be even potentially rewarding are more restrictive than might be expected, and power improvements are generally poor. Using only elementary arguments and calculations with popular spreadsheets, analyses of standard simple cases show two in which pooling is possibly both sensible and of practical potential benefit: Pooling across sources and pooling of highest and next-to-highest order interaction terms as an error term in the unreplicated 2ndesign case. The special case of paired or unpairedttests is also examined, and the sample correlation needed to make the pairedttest more powerful than the unpaired one is found to be both low and well below conventionally statistically significant values.
点击下载:
PDF (1452KB)
返 回