A0264
Title: Kalman filter for innovations state space models
Authors: Chiu-Hsing Weng - National Chengchi University (Taiwan) [presenting]
Chan-Yuan Hsu - National Chengchi Univ (Taiwan)
Abstract: Exponential smoothing methods have been widely used for time series forecasting. To calculate likelihood and prediction intervals, many have specified various state space models that underlie exponential smoothing methods. Among these state space models, the innovations formulation, called innovations state space models, is a popular one. With a state space representation, one can easily incorporate regressors in exponential smoothing methods. We derive Kalman filter type algorithms for innovations state space models with and without regressors, where the regressors can be fixed or random. Some examples are used for illustration.
