Diagnostics for multiple regression problems [electronic resource].
- Richland, Wash. : Pacific Northwest Laboratory, 1982.
Oak Ridge, Tenn. : Distributed by the Office of Scientific and Technical Information, U.S. Dept. of Energy.
- Additional Creators:
- Pacific Northwest Laboratory
United States. Department of Energy. Office of Scientific and Technical Information
- In the last 10 to 15 years there has been much work done in trying to improve linear regression results. Individuals have analyzed the susceptibility of least-squares results to values far removed from the center of the independent variable observations. They have studied the problem of heavy-tailed residuals, and they have studied the problem of collinearity. From these studies have come ridge regression techniques, robust regression techniques, regression on principal components, etc. However, many practitioners view these methods with suspicion (and ignorance), and prefer to continue using the usual least-squares procedures to fit their models, even though their results might not be answering the question they think. In reaction to this, statisticians are spending more time analyzing how the individual observations affect the least squares results. In the last few years approximately 10 papers and one text have appeared that address the problem of how to study the influence of the individual observations. This report is a study of the recent work done in linear regression diagnostics. It is concerned with analyzing the effect of one case at a time, since the methods to analyze this situation are relatively straight-forward and are not prohibitive computationally.
- Published through SciTech Connect.
- Funding Information:
View MARC record | catkey: 14357482