CLASSICAL DESIGN OF EXPERIMENTS APPROACH MAY BE NON- EFFECTIVE FOR ACTIVE IDENTIFICATION PROBLEMS

摘要

The classical design of experiments approach is used in problems of active identification very intensive (see, for example, [l]-[5]). Here the classical optimum designs of experiments are investigated. Extending these results, methods are obtained for investigating of the optimum designs in various settings. Many examples are given where optimal design and non-optimum one are compared. The execution of the equality in equivalence theorem is considered and investigated. The main result is so. A distance between a prior dispersion function and a posterior one can be very essential and the equality in Kiefer's theorem (see [6]-[10]) can be not true for a posterior data.