Intervalos de previsão bootstrap para modelos estruturais

• Ano de defesa: 2011
• Autor(a) principal: Thaize Vieira Martins
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: sazonalidade; Forma de espaço de estados; filtro de Kalman; bootstrap; Bootstrap (Estatística); Estatística; Analise multivariada; Series temporais; Análise espacial (Estatística)
• Link de acesso: http://hdl.handle.net/1843/ICED-8HNHU8

This work is dedicated to the implementation of the methodology to calculate nonparametric bootstrap prediction intervals in state space form (SS), based on the work of Rodriguez and Ruiz (2009). The SS are an alternative way of rewriting the structural models, which decompose the time series in their non-observable components (level, trend and seasonality). Mainly, this work has the interest of extending the methodology proposed by Rodriguez and Ruiz (2009) to more complex structural models, implementing the algorithms in the Ox language and comparing the simulation results with the traditional method of constructing asymptotic prediction intervals and also with the parametric bootstrap procedure.