Title: Efficient estimation for non-linear state space models of population survey data
Authors: Takis Besbeas - Athens University of Economics and Business (Greece) [presenting]
Abstract: Time series data of population abundances are often described using population dynamics state-space models involving Gompertz, Moran-Ricker or Beverton-Holt latent processes. We show how hidden Markov model methodology provides a flexible framework for fitting a wide range of models to such data. The proposed method avoids any Kalman filter approximations or Monte Carlo simulation that might be employed, and allows model comparison and goodness-of-fit using standard likelihood tools. The method is illustrated using two real data sets of mammal populations from Europe and Australia. There is little difference between the three latent models for the two case studies, which suggests ecological time series may not be sufficiently informative on latent structure when observation error is unknown in general.