Estimação por maxima quase-verossimilhança no dominio do tempo de modelos de volatilidade estocastica com memoria longa

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

The aim of this work is to use of the Quasi-Maximum Likelihood estimator of Long Memory Stochastic Volatility models. The estimation by quasi-maximum likelihood is done in the time domain using an approximate state-space representation of the model. We analyse the autoregressive and moving average approximation using deferent maximum lag approximation. The results are compared to usual frequency domain estimators proposed by Geweke and Porter-Hudak and Robinson applied to ARFIMA models. The methods are applied to simulated series of sample size 1000 and to the São Paulo Stock Exchange Index (IBOVESPA) series. The series was collected at one-minute and five-minute intervals, from december 18th, 2001 to January 2sd, 2002.

ASSUNTO(S)

estimativa de parametro analise de series temporais verossimilhança (estatistica)

Documentos Relacionados