Identification in Control and Econometrics; Similarities and Differences.
Interim technical rept.,
HARVARD UNIV CAMBRIDGE MASS DIV OF ENGINEERING AND APPLIED PHYSICS
Pagination or Media Count:
The report attempts to bridge the gap between the economic and the control literatures on the subject of system identification and parameter estimation. It is pointed out that the emphasis in the economic literature is on large simultaneous equation models and linear estimation techniques, whereas the emphasis in the control literature is on state vector and transfer function models, on problems due to partial state observations and nonlinear estimation techniques. Since a step in the direction of easier communication between researchers in the two fields would be the use of a common model, the state-vector model of control which has already been used in several economic studies is proposed as a unifying link. The relationship of the state vector model to the simultaneous equation model and the role of process and measurement noise in the econometric context are discussed. Complete results on the identifiability of state-vector models along with a stepwise two-stage least squares method for model structure determination and a maximum likelihood method for parameter estimation are given. The problems of closed-loop system identification and input design are also briefly discussed. Author
- Economics and Cost Analysis
- Statistics and Probability