TAR Modeling with Missing Data when the White Noise Process Follows a Student's t-Distribution

Abstract

This paper considers the modeling of the threshold autoregressive (TAR) process, which is driven by a noise process that follows a Students t-distribution. The analysis is done in the presence of missing data in both the threshold process {Zt} and the interest process {Xt}. We develop a three-stage procedure based on the Gibbs sampler in order to identify and estimate the model. Additionally, the estimation of the missing data and the forecasting procedure are provided. The proposed methodology is illustrated with simulated and real-life data.

Keywords

• datos faltantes
• estadística Bayesiana
• modelo autoregresivo de umbrales
• muestreador de Gibbs
• pronóstico
• series de tiempo