Numerically Stable Methods for Updating Regressions

Publisher Website
Google Scholar
Abstract
A stable numerical procedure is presented for updating regressions, including quantities used in the analysis of variance, when regressors or observations are added or deleted. The modified Gram-Schmidt triangular factorization is used, and the matrix with orthonormal columns is stored and updated. In this way it is possible to monitor and to minimize the effects of rounding errors through the techniques of iterative refinement and reorthogonalization. Numerical results are included.

This publication has 0 references indexed in Scilit: