Estimating linear dynamical models using subspace methods
Dietmar Bauer
Institut for Econometrics, Operations Research and System Theory
TU Wien, Austria
Abstract: This talk presents a survey on a special class of subspace
methods.
These methods are alternative to prediction error methods for the estimation
of linear dynamical models of the ARMA type (in the state space framework)
fitted to time series data. The main advantages of subspace methods
lie in
their conceptual simplicity and the lower computational load as compared
to
prediction error methods. The basic underlying idea is the exploitation
of the
properties of the state.
The talk includes a detailed description of the methods followed by
a
discussion of the results on the asymptotic properties of the algorithm.
The
main emphasis here will be on explicit expressions for the asymptotic
variance.
Seminar to be held in Room 107, 24 Hillhouse at 4:15 pm