The maximum likelihood estimation procedure for estimation of parameters in nonlinear state space (NLSS) models is studied. In order to achieve a more accurate and robust estimation, the Monte Carlo expectation maximum (MCEM) algorithm is implemented using a particle independent Metropolis-Hastings (PIMH) procedure. An automated sample size selection criterion is constructed based on renewal theory to determine whether the sample size needs to be increased at each iteration in implementation of the MCEM method with the PIMH algorithm. To illustrate the proposed methodology, an application to the stochastic volatility model is considered for which a simulation study is conducted. An empirical study of daily exchange rates based on the NLSS model also demonstrates application and performance of the proposed approach. Results show the proposed procedure provides better estimation performance. (22 refs.).
展开▼
