Ciclo de Palestras do PPG em Estatística – IM-UFRJ
Data: 08 de Maio de 2015
Hora: 15:30 h
Local: LSE – Laboratório de Sistemas Estocásticos, sala I-044b
Palestrante: Cibele Queiroz da Silva (UNB)
Título: Th Dynamic Dirichlet Model
Resumo: In this talk we present a new dynamic model, the Dynamic Dirichlet Model (DDM), for describing time series of compositional data. Such kind of data is characterized by random vectors yt defined on the open standard (k-1)-simplex. Each coordinate of yt represents the share, in percentage, of each one of the k categories that describe a given phenomena. The DDM includes, as sub-models, the Beta Dynamic Model (da-Silva, et al., 2011- CSDA), the static Dirichlet regression and the static Beta regression (Ferrari and Cribari-Neto, 2004-JAS). We designed both on-line and off- line approaches for the estimation of the parameters in the model. The on- line version is adequate for recursive estimation while the off-line one, which is based on stochastic simulation via MCMC, can be used when there are some specific unknown parameters in the model. We discuss the practical use of the proposed model in describing the past behaviour of the series, as well as in the prediction process.
Contato: (21) 3938-7374