An adaptive resampling scheme for cycle estimation

Resumo

Bayesian dynamic linear models (DLM) are useful in time series modeling because of the flexibility that they present in obtaining a good forecast. They are based on a decomposition of the relevant factors which explain the behavior of the series through a series of state parameters. Nevertheless the DLM as developed in West & Harrison (1997) depend on additional quantities, such as the variance of the system disturbances, which, in practice, are unknown. These are refered as hyperparameters of the model. In this paper, DLM with auto-regressive components are used to describe time series showing cyclic behavior. The marginal posterior distribution for state parameters can be obtained by weighing the conditional distribution of state parameters by the marginal distribution of hyperparameters. In most cases the joint distribution of the hyperparameters can be obtained analytically but the marginal distributions of the components can not, thus requiring numerical integration. We propose to obtain samples of the hyperparameters by a variant of the Sampling Importance Resampling (SIR) method. A few applications are made with simulated and real datasets.

Ficha do documento

Tipo

Estudo

Ano

1998

Instituição

Instituto de Pesquisa Econômica Aplicada (Ipea)

Fonte

Repositório do Ipea

Idioma

Inglês

Acesso

Acesso aberto

Identificador

oai:repositorio.ipea.gov.br:11058/2143

Licença

Licença Comum

Palavras-chave

Modelos lineares dinâmicos (MLD)Previsão de curto prazo de séries de tempo