Analise de componentes ciclicas em series temporais uni e multivariadas via filtros HP modificados e outros metodos / Cycle components analysis for uni and multivariate time throug modified Hodrick-Prescott filters and other methods
AUTOR(ES)
Jaqueline Barbão
DATA DE PUBLICAÇÃO
2007
RESUMO
The analysis of cycles in time series is considered in this investigation through four methodologies, of which three were more recently developed in literature. The first method, proposed in Kaiser and Maravall (2001), is based on a modification of the unidimensional Hodrick-Prescot filter (HP) through ARIMA models approach. The second method, developed in Mills (2003), consists of a multivaxiate extension of the HP filter, whereas the third method, which follows Durbin and Koopman (2001), decomposes the series through either univariate or multiariate state space models. The fourth method, which is traditional in literature and based on harmonic non-linear regression models, is taken in order to be used as standard reference for comparison. These methods are applied in three Spanish macroeconomic time series (industry production index, cement consumption and airline tickets sales) and two natural phenomena time series (rainfall in Fortaleza/Brazil and sunspot average number). These methodologies are examined with respect the facilities of the implementation, forecasting and description capability, and the quality of the estimated cycle component
ASSUNTO(S)
chuvas - periodicidade series temporais manchas solares time series rain and rainfall cycles sunspot numbers
ACESSO AO ARTIGO
http://libdigi.unicamp.br/document/?code=000393041Documentos Relacionados
- Tendência na incidência de acidentes e doenças de trabalho no Brasil: aplicação do filtro Hodrick-Prescott
- ANÁLISES CLIMÁTICAS: O FILTRO HODRICK-PRESCOTT APLICADO AOS ÍNDICES ATMOSFÉRICOS DA OSCILAÇÃO SUL E DA OSCILAÇÃO DO ATLÂNTICO NORTE
- O uso de quase U-estatísticas para séries temporais uni e multivaridas
- Métodos de reamostragem de séries temporais baseados em wavelets.
- Redes Bayesianas: um método para avaliação de interdependência e contágio em séries temporais multivariadas