BAYESIAN ANALYSIS OF DYNAMIC FACTOR MODELS USING MULTIVARIATE T DISTRIBUTION

Main Article Content

Larissa Ribeiro de ANDRADE
Daniel Furtado FERREIRA
Thelma SÁFADI
Lúcia Pereira BARROSO

Abstract

The multivariate t models are symmetric and have heavier tail than the normal distribution and produce robust inference procedures for applications. In this paper, the Bayesian estimation of a dynamic factor model is presented, where the factors follow a multivariate autoregressive model, using the multivariate t distribution. Since the multivariate t distribution is complex, it was represented in this work as a mix of the multivariate normal distribution and a square root of a chi-square distribution. This method allowed the complete dene of all the posterior distributions. The inference on the parameters was made taking a sample of the posterior distribution through a Gibbs Sampler. The convergence was veried through graphical analysis and the convergence diagnostics of Geweke (1992) and Raftery and Lewis (1992).

Article Details

How to Cite
ANDRADE, L. R. de, FERREIRA, D. F., SÁFADI, T., & BARROSO, L. P. (2018). BAYESIAN ANALYSIS OF DYNAMIC FACTOR MODELS USING MULTIVARIATE T DISTRIBUTION. Brazilian Journal of Biometrics, 36(1), 140–156. https://doi.org/10.28951/rbb.v36i1.155
Section
Articles