Análise de intervenção via estimação clássica e Bayesiana de fatores de descontouma aplicação para o índice da produção industrial no Brasil

Resumo

It is specified an univariate structural model, that decomposes time series in nonobservable components, for the Brazilian' s index of industrial production. We compare the estimates for the trend and seasonal factors when two different estimation methods, of the discount factors, are used: a classical and a Bayesian (Sampling and Re-sampIing) method. To detect periods with structural breaks we monitor the series using the bayesian procedure suggested by West (1988). The interventions, for the periods with structural breaks, are carried on imposing a different discount factor for these periods. The main results are: a) the trend and seasonal components of the index of industrial production are not significantly different when the classical or Bayesian methods of estimation of the discount factors are used. This results depends heavily on the shape of the Iikelihood that has a peak in a small region of the set of possible values for the discounts; b) fixing subjectively the discount factors can result in a substancial departure of the prior distribution of the discount factors from the likelihood; c) the estimated seasonal factors are not very different from those obtained employing the X Il-Arima method.

Ficha do documento

Tipo

Estudo

Ano

1997

Instituição

Instituto de Pesquisa Econômica Aplicada (Ipea)

Fonte

Repositório do Ipea

Idioma

Português

Acesso

Acesso aberto

Identificador

oai:repositorio.ipea.gov.br:11058/2163

Licença

Licença Padrão Ipea

Abrangência

Brasil

Palavras-chave

Estimação clássica e bayesiana de fatores de descontoÍndice da produção industrialModelo univariado estrutural