CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: The local partial autocorrelation function and its application to forecasting Authors:  Marina Knight - University of York (United Kingdom) [presenting]
Rebecca Killick - Lancaster University (United Kingdom)
Guy Nason - Imperial College, London (United Kingdom)
Idris Eckley - Lancaster University (United Kingdom)
Abstract: The regular and partial autocorrelation functions are powerful tools for stationary time series modelling and analysis. However, in many applied situations time series are not stationary and in these settings the use of the regular (classical) and partial autocorrelations can give misleading answers. We introduce the local partial autocorrelation function and establish the asymptotic behaviour of its estimators. We demonstrate its practical utility as a tool in economic applications and in addition, we propose its use to improve the forecasting of locally stationary time series. Our new forecasting method exhibits excellent forecasting and prediction interval coverage for simulated nonstationary data, as well as for economic and wind energy data.