I've been working developing a least squares routine that makes use of the drivers GELS,GELSY,GELSS,and GELSD.
I think I've been able to work out how to use the "A" matrix on output to estimate the covariance matrix for all but GELSD.
Unfortunately, the doc for GELSD is simply "On exit, A has been overwritten." Does anyone have some suggestions for
How to estimate the covariance matrix from the output of DGELSD(F77) or GELSD(F95). It doesn't seem efficient or consistent
to calculate (A^T*A) brute force after GELSD has gone through the trouble of rescaling and transforming the matrix for accuracy and stability considerations.
Any Suggestions would be welcome,