On Consistency and Identifiability
1976 (English)In: Stochastic Systems: Modeling, Identification, and Optimization / [ed] Roger J.-B. Wets, Springer Berlin/Heidelberg, 1976, 169-190 p.Chapter in book (Refereed)
The convergence with probability one, of the parameter estimates obtained from prediction error identification methods, such as the maximum likelihood method, is analysed in this paper. It is shown that under quite weak assumptions on the actual system that has generated the data, the expected value of the identification criterion can be used for the asymptotic analysis of the estimates. In particular, the true system does not have to belong to the set of models over which the search for optimum is made. The implications of this result for consistency analysis and for questions of identifiability, as well as for other related problems are discussed.
Place, publisher, year, edition, pages
Springer Berlin/Heidelberg, 1976. 169-190 p.
, Mathematical Programming Studies, ISSN 0303-3929 ; 5
Convergence, Parameter estimation, Error identification, Maximum likelihood
IdentifiersURN: urn:nbn:se:liu:diva-100810DOI: 10.1007/BFb0120772ISBN: 0-7204-0569-6OAI: oai:DiVA.org:liu-100810DiVA: diva2:663742