Information, Consistent Estimation and Dynamic System Identification.
MASSACHUSETTS INST OF TECH CAMBRIDGE ELECTRONIC SYSTEMS LAB
Pagination or Media Count:
The asymptotic behavior of parameter estimates and the identification and modeling of dynamical systems are investigated. Measures of the relevant information in a given sequence of observations are defined and shown to possess useful properties, such as the metric property on the parameter set. The convergence of maximum likelihood and related Bayesian estimates for general observation sequences is investigated. The situation where the true parameter is not a member of a given parameter set is considered as well as the situation where the parameter set includes the true model. The finite parameter set case is emphasized for simplicity in the convergence analysis, but the results are extended in general terms to the infinite parameter case.
- Operations Research