Analysis by Computer of Designed Experiments with Zero Degrees of Freedom for Error Variance
作者:
WilliamH. Collins,
CarolB. Collins,
期刊:
The American Statistician
(Taylor Available online 1989)
卷期:
Volume 43,
issue 3
页码: 165-168
ISSN:0003-1305
年代: 1989
DOI:10.1080/00031305.1989.10475648
出版商: Taylor & Francis Group
关键词: Analysis of variance;External value for mean squared error;Regression;Saturated design;Statistical computing
数据来源: Taylor
摘要:
Some statistical software refuses to perform regression and analysis of variance on data sets whose models allow no degrees of freedom for error variance. A related problem is getting statistical software to use a particular value for mean squared error for significance tests. We show how an external estimate of error variance can be entered via fake data in designed experiments so that the software works and the effects studied are not changed.
点击下载:
PDF (397KB)
返 回