A1062
Title: Nonparametric estimation and forecasting of structural time series models
Authors: Joachim Schnurbus - University of Passau (Germany) [presenting]
Harry Haupt - University of Passau (Germany)
Abstract: A non-additive nonparametric approach is proposed for estimation and forecasting of structural time series models containing trend and seasonal components as well as additional covariates. A new kernel function is proposed that allows to take into account the specific ordered structure of seasonal effects. We provide a Monte Carlo analysis of the forecasting performance against competitors such as seasonal ARIMA and nonlinear innovation state-space models.