A0781
Title: Computationally efficient forecasting algorithm in the SUTSE model and its properties
Authors: Wataru Yoshida - Kyushu University (Japan) [presenting]
Kei Hirose - Kyushu University (Japan)
Abstract: The problem of forecasting multivariate time series is considered by using a Seemingly Unrelated Time Series Equations (SUTSE) model. In the SUTSE model, multiple univariate time series equations are combined to express a single state-space model, resulting in the coefficient matrices for system and observation models to be diagonal. The SUTSE model usually assumes the correlations among error variables. In this case, however, the model estimation requires heavy computational loads due to a large matrix computation, especially for high-dimensional data. To alleviate the computational issue, we propose a two-stage procedure for forecasting. First, we conduct the Kalman filter as if the correlations among error variables do not exist; that is, univariate time series analyses are conducted separately. Next, the forecast value is computed by estimating a covariance matrix of the forecast error. The proposed algorithm is much faster than the ordinary SUTSE model because we do not require a large matrix computation. Some theoretical properties of our proposed estimator are presented. Monte Carlo simulation is conducted to investigate the effectiveness of our proposed method. The results show that our proposed method is comparable with the ordinary SUTSE in prediction accuracy.
