"FORECASTING NON-STATIONARY TIME SERIES BY WAVELET PROCESS MODELLING"

This page accompanies the article Forecasting non-stationary time series by wavelet process modelling by Piotr Fryzlewicz, Sebastien Van Bellegem and Rainer von Sachs. The reader can download here the adaptive forecasting routines written for the paper, as well as the data sets analysed in it.

All the functions have been written in S and designed for S-Plus (version "2000 Windows"), which is a commercial package. However, they will also probably work in R, which can be downloaded free of charge here.

Downloading and installing the functions and the data sets.

  1. Click here to download the archive "prediction.zip", which contains the functions and the data sets.
  2. Extract the archive. You will obtain the file "prediction.q".
  3. Run S-Plus and type (in Windows) "data.restore("D:\\PATH\\prediction.q")", where D:\PATH is the directory containing "prediction.q".
  4. The functions and the data sets are now available.
Also, the functions (but not the data sets!) are available here in a text format.

Alphabetic list of functions --- click on the name for help

Alphabetic list of data sets --- click on the name for description