Numerical Computations for Univariate Linear Models
Abstract
The authors consider the usual univariate linear model E(y) = Xgamma, V(y) = Sigma squared. In part one of the paper X has full column rank. Numerically stable and efficient computational procedures are developed for the least squares estimation of gamma and the error sum of squares. An orthogonal triangular decomposition of X is used, using Householder transformations. A lower bound for the condition number of X is immediately obtained from this decomposition. In part two, X has less than full rank. Least squares estimates are obtained using generalized inverses. The function L'gamma is estimable whenever it admits an unbiased esimator linear in y.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1971
- Accession Number
- AD0737648
Entities
People
- Gene H. Golub
- George P. H. Styan
Organizations
- Stanford University