An example of propagation of error associated with an iterated least squares solution (Q582030)

The author considers the iterative least squares solution of the covariance matrix related to the nonlinear problem: \(f(x)=b\), linearized into: \(A(x)[x-x(0)]=b-b(0),\) \(A\underline\triangle \partial f/\partial x.\) It is shown that the covariance matrix is: \(\sigma_ x'=\sigma_ x+C\), where \(\sigma_ x\) is the pseudo inverse of the covariance of the measurements b, and C a matrix depending from the dependence of A on b. Examples are provided.