The extended classical total least squares algorithm (Q1115101)

The authors describe a FORTRAN subroutine for an extended version of the classical total least squares problem [cf. \textit{H. Golub} and \textit{C. F. Van Loan}, SIAM J. Numer. Anal. 17, 883-893 (1980; Zbl 0468.65011)]. The algorithm solves the overdetermined system \(AX=B\) where each of the observation matrix A and the data matrix B are inaccurate. It is assumed that errors in [A;B] are independent and of equal size; this implies that the user has performed all preprocessing, such as scaling and co-ordinate transformations, before calling the subroutine. The subroutine is also able to solve determined and under-determined systems by computation of the minimum norm solution.