Taylor, James (2010) Reply to the discussion of: Exponentially weighted methods for forecasting intraday time series with multiple seasonal cycles. International Journal of Forecasting, 26 (4). pp. 658-660.
Alysha De Livera’s discussion focuses on the transformation of the data and the inclusion of ARMA terms for the residuals of the exponential smoothing models. Drawing on the formulation for the HWT model, I implemented all exponential smoothing models with an AR(1) term included for the residual. In my paper, I applied a logarithmic transformation to the NHS Direct data prior to fitting all methods. The BATS model, presented in Alysha De Livera’s discussion, is a generalisation of the implementation of the HWT model in my paper, with a Box-Cox transformation and ARMA terms of different lags for the residual.
|Keywords:||Seasonality; Intraday data; Call centre arrivals; Exponential smoothing; Exponential weighting; Discount weighted regression; Regression splines; Singular value decomposition|
|Date Deposited:||24 Apr 2012 15:09|
|Last Modified:||10 Feb 2017 16:35|
Actions (login required)