B0617
Title: A flexible regression model for functional time series
Authors: HyeYoung Maeng - London School of Economics (United Kingdom) [presenting]
Piotr Fryzlewicz - London School of Economics (United Kingdom)
Abstract: A prediction model is introduced for functional time series. Based on the classical scalar-on-function regression, the idea is to split the observed daily curves into several pieces to apply different smoothness. The proposed model allows more smoothing on observations located far from the prediction point compared to the closely located ones. In contrast to the typical time series prediction, our proposal gives flexibility in modelling in the sense that it offers less weight on the interval which is considered less important than others by fitting functional variables. The model in its simplest form includes one functional and one or more scalar covariates which is classified as semi-functional regression. In our approach, the change point which divides scalar and functional variable can be estimated from data. Illustrations on real data sets are given to show that the new model outperforms existing competitors. The asymptotic properties will also be presented.
Journal