Bootstrap-based estimates of uncertainty in subspace identification methods (Q5925914)

The authors study the problem of evaluation of model uncertainty in the framework of subspace identification methods. They use methods of computational statistics. In particular, a procedure for the estimation of the standard error of the eigenvalues and of the frequency response of the identified model is presented. The statistical tool used is the bootstrap method. Simulated results are given and demonstrate that accurate estimates of the standard error are obtained by this technique.