Abstract
Based on permutations and scaling of orthogonal Hessenberg matrices a family of algorithms (the G- and partial G-algorithms) are presented which solve the weighted least squares problem (Ax−b)’W(Ax−b)=ρ2 for x such that ρ2 is minimal. A sequence of G-transformations triangularizes A and yields ρ2 as a by-product quite similar to the Householder-Golub algorithm for the equiweighted problem, but the transformed matrices are sparser, and no square roots are required. A comparison of the numerical results obtained by the Basic G-Algorithm with those obtained by the Householder, modified Gram-Schmidt, LU-factorization (and the normal equation) methods indicates that the new method performs as well as the best available methods, and is more flexible with respect to different types of applications. Also the unscaled covariance matrix (A’WA)−1 is efficiently obtained after A has been subjected to the G-transformations.
The Fast Givens Method is a special form of the partial G-transform. The G-transformations have many more applications than presented here. The theory of the G-transformations closes a gap left open in the mathematical development of the linear least squares problem.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Householder, Alston S. (1953) Principles of numerical analysis. McGraw Hill Book Co. N. Y.
Businger, P. and Golub, G. H. (1965) Linear least squares solutions by Householder transformations. Numer. Math. 7, 269–276
Rice, J. R. (1966) Experiments on Gram-Schmidt oethogonalization. Math. Comp. 20, 325–328
Pereyra, V. (1968) Stabilizing linear least squares problems. 68, Proc. IFIP Congress 1968
Peters, G. and Wilkinson, J. H. (1970) The least squares problem and pseudo-inverses. Computer J. 13, 309–316
Björck, Åke (1970) Methods for sparse linear least squares problems, in Sparse matrix computations, J. R Bunch & D. J. Rose, Eds. Academic Press, N. Y.
Gentleman, W. M. (1973) Least squares computations by Givens transformations without square roots. J. Inst. Math. Appl. 12, 329–336
Dahlquist, G. and Björck, Å. (1974) Numerical Methods. Prentice Hall, Inc., New Jersey
Golub, G. H. and Styan, G. P. (1973) Numerical computations for univariate linear models. J. Stat. Comput. Simul., 2, 253–274
Gill, P. E., Golub, G. H., Murray, W., and Saunders, M. A. (1974) Methods for modifying matrix factorizations. Math. Comp. 28, 505–535
Daniel, J. W., Gragg, W. B., Kaufman, L., and Stewart G. W. (1976) Reorthogonalization and stable algorithms for updating the Gram-Schmidt QR-factorization. Math. Comp. 30, 772–795
Lawson, C. L. and Hanson, R. J. (1974) Solving least squares problems. Prentice Hall, N.J.
Seber, G. A. F. (1977) Linear regression analysis. Wiley, N.Y.
Hansen, J. E. (1981) Testing and evaluation of the G-algorithm for solving weighted least squares problems. MS Thesis, Northwestern University, Evanston, IL 60201
Bareiss, E. H. (1981) Matrices and Tensors, in Lecture notes in numerical analysis. Northwestern University, Evanston, IL 60201. 82-01-NAM-01
Bareiss, E. H. (1981) Least squares solution of linear systems by G-and H-transformations. Appendix H in Hansen (1981), 127–163
Bareiss, E. H. (1982) Numerical solution of the weighted linear least squares problem by G-transformations. Northwestern University, Evanston, IL 60201. 82-03-NAM-03
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1983 Springer-Verlag
About this paper
Cite this paper
Bareiss, E.H. (1983). Least squares solution of weighted linear systems by G-transformations. In: Numerical Methods. Lecture Notes in Mathematics, vol 1005. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0112522
Download citation
DOI: https://doi.org/10.1007/BFb0112522
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-12334-7
Online ISBN: 978-3-540-40967-0
eBook Packages: Springer Book Archive