ABSTRACT
As we saw in Section 11.1.3, two types of inferential problems arise from the nonlinear dynamic state space models:
(a) The estimation problem of the unobserved state variables from the observed time series (possibly contaminated with observation errors) assuming that the given state space model is true
(b) The identi–cation problem of the model of the state equation (i.e., dynamical system model) from the partially observed time series, where some of the variables of the state equation are not observed, and the observed variables are possibly contaminated by observation errors
Here in problem (b), the true state space model is unknown, and we have 2 possible cases:
(b − 1) The model structure (i.e., the dimension of the state and parametric form specifying the state space model) is known, but precise parameter values are unknown.
