You are not currently logged in.
Access your personal account or get JSTOR access through your library or other institution:
If You Use a Screen ReaderThis content is available through Read Online (Free) program, which relies on page scans. Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
The Use of Guided Reformulations when Collinearities are Present in Non- Linear Regression
Jeffrey S. Simonoff and Chih-Ling Tsai
Journal of the Royal Statistical Society. Series C (Applied Statistics)
Vol. 38, No. 1 (1989), pp. 115-126
Stable URL: http://www.jstor.org/stable/2347686
Page Count: 12
Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Preview not available
Multicollinearity is as serious a problem in non-linear regression as it is in linear regression, with the vectors of partial derivatives taking the place of the columns of the design matrix. The solution to this problem is to reformulate the model, but this is often difficult. In this paper a guided reformulation procedure is developed based on the representation of collinearity in the model. Several examples are treated, and it is shown that the reformulation model generally has reduced collinearity and curvature with little loss of fit. A discussion of the relationship of collinearity to curvature for a certain class of non-linear models is provided. Potential difficulties with the procedure are also discussed.
Journal of the Royal Statistical Society. Series C (Applied Statistics) © 1989 Royal Statistical Society