A0467
Title: Nonstationary functional time series forecasting
Authors: Yang Yang - University of Newcastle (Australia) [presenting]
Han Lin Shang - Australian National University (Australia)
Abstract: A nonstationary functional time series forecasting method is proposed with an application to age-specific mortality rates observed over the years. The method begins by taking the first-order differencing and estimates its long-run covariance function. Through eigen-decomposition, a set of estimated functional principal components and their associated scores are obtained for the differenced series. These components allow the reconstruction of the original functional data and compute the residuals. To model the temporal patterns in the residuals, dynamic functional principal component analysis is again performed, and its estimated principal components and the associated scores for the residuals are extracted. As a byproduct, a geometrically decaying weighted approach is introduced to assign higher weights to the most recent data than those from the distant past. Using the Swedish age-specific mortality rates from 1751 to 2022, the weighted dynamic functional factor model is demonstrated to produce more accurate point and interval forecasts, particularly for male series exhibiting higher volatility.
