Generalized QR factorization and its applications (Q1183190)

The concept of generalized \(QR\) factorization (\(GQR\) factorization) considered earlier by S. Hammarling (1986) and C. Paige (1990) is developed. The terminology \(GQR\) factorization refers to the orthogonal transformations that simultaneously transform two matrices \(A\) and \(B\), having the same number of rows, to triangular form. The authors present a more general form of the factorization that relaxes the requirements on the rank of some submatrices in the factored form. It simplifies the development of software and does not limit the class of application problems that can be solved. The \(GQR\) factorization provides the uniform approach to solve the linear equality-constrained least squares problem, and the generalized linear regression model. The conditioning of these problems is estimated and \(GQR\) factorizations with pivoting and without pivoting are distinguished.