Estimation for Parameter-Driven State-Space Models
by
Richrad A. Davis
Department of Statistics, Colorado State University
Coauthors: Gabriel Rodriguez-Yam

Typically, the likelihood function for non-Gaussian state-space models can not be computed explicitly and so simulation based procedures, such as importance sampling or MCMC, are commonly used to estimate model parameters. In this paper, we consider an alternative estimation procedure which is based on an approximation to the likelihood function. The approximation can be computed and maximized directly, resulting in a quick estimation procedure without resorting to simulation. Moreover, this procedure is competitive with estimates produced using simulation-based procedures. The speed of this approach makes it viable to fit a wide range of potential models to the data and allows for bootstrapping the parameter estimates. This procedure will be illustrated in two examples; fitting a stochastic volatility model to a time series of exchange rates and fitting a Poisson model to a time series of counts.

Date received: July 31, 2003