Bayesian Copula Factor Autoregressive Models for Time Series Mixed Data - Hadi Safari Katesari
This event is in the past.
Hadi Safari Katesari
School of Mathematical and Statistical Sciences, Southern Illinois University Title:
Bayesian Copula Factor Autoregressive Models for Time Series Mixed Data
Abstract: In this talk, we propose a Bayesian copula factor Autoregressive models for time series mixed data. This is a novel model that assumes conditional independence and applies latent factors in both response time series and in high-dimensional mixed-type covariates of the time series with the quadratic regression model. The framework of the model gives an efficient dimension reduction and characterizes the main effects and interactions of the covariates by including the latent variables in the response time series. We apply a semiparametric time series extended rank likelihood for the margins of explanatory variables which reduces the number of estimated parameters and provides a fast computational algorithm. A flexible Bayesian algorithm is proposed to compute the posterior distribution of latent factors and model parameters with Metropolis-Hastings and Forward Filtering Backward methods within Gibbs sampling. The theoretical results and MCMC computations are evaluated with simulation studies. Moreover, the proposed model is applied to the quarterly U.S. economic dataset.