JOINT CONFIDENCE REGIONS FOR MULTIPLE REGRESSION COEFFICIENTS
成果类型：
Article
署名作者：
DURAND, D
刊物名称：
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
ISSN/ISSBN：
0162-1459
DOI：
10.2307/2281040
发表日期：
1954
页码：
130-146
关键词：
摘要：
In multiple regression problems two popular practices involve fallacies. One is the procedure of eliminating terms whose estimated regression coefficients do not differ significantly from zero. The other is making confidence interval statements about individual coefficients with the implication that the associated probability applies simultaneously to all coefficients. With the purpose in mind of avoiding these fallacies, the author discusses methods of making joint confidence interval statements with a single over-all confidence probability and associated methods of deriving statements about subsets of coefficients. He also gives lower bounds for the probability that all conventional confidence intervals will be correct. Illustrative examples are given.
来源URL：