EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0513
Title: An alternative estimation of a time-varying parameter model Authors:  Tatsuma Wada - Keio University (Japan) [presenting]
Mikio Ito - Keio University (Japan)
Akihiko Noda - Kyoto Sangyo University (Japan)
Abstract: A non-Bayesian, generalized least squares (GLS) based approach is proposed to estimate a class of time-varying AR parameter models. This approach is proven very efficient because unlike conventional methods, it does not require the Kalman filtering and smoothing procedures, but yields the smoothed estimate that is identical to the Kalman smoothed estimate. Since the GLS based approach can compute the smoothed estimate at once, the time required to compute the smoothed estimate is shorter than the time required by alternative approaches. In addition, this approach enables us to deal with stochastic volatility models and models having a time-dependent variance-covariance matrix. Other features, such as the possibility of the pile-up problem are assessed through simulations.