Inferência sobre os hiperparâmetros dos modelos estruturais usando Bootstrap

This dissertation is based on the decomposition of times series via non-observed components, through structural models. An alternative way to rewrite the structural models is by using the state space form. Once this transcription is done, the Kalman filter is used for updating the state vector and c...

ver descrição completa

Detalhes bibliográficos
Autor: Juliana Aparecida Ribeiro
Formato: tesis de maestría
Estado:Versión publicada
Fecha de publicación:2006
País:Brasil
Recursos:Universidade Federal de Minas Gerais (UFMG)
Repositorio:Repositório Institucional da UFMG
Idioma:portugués
OAI Identifier:oai:repositorio.ufmg.br:1843/RFFO-7HPSWM
Acesso em linha:http://hdl.handle.net/1843/RFFO-7HPSWM
Access Level:acceso abierto
Palavra-chave:Inferencia
Modelos
Bootstrap (Estatística)
Estatística
Kalman, Filtragem
Series temporais
Inferencia (Logica)
Probabilidades
Verossimilhança (Estatística)
Teoria da estimativa
Descrição
Resumo:This dissertation is based on the decomposition of times series via non-observed components, through structural models. An alternative way to rewrite the structural models is by using the state space form. Once this transcription is done, the Kalman filter is used for updating the state vector and constructing the likelihood function to estimate the hyperparameters of the model. The bootstrap resampling technique is applied to make inferences on the hyperparameters of the models, attending for the construction of confidence intervals, which is made in the programming language Ox. The results of the simulations and a real time series application verify the efficiency of the language estimation process.Keywords: structural models, Kalman filter, hyperparameters, bootstrap.