Re: Comparing Forecasting Models

dave_at_autobox.com
Date: 01/21/05 

Date: 21 Jan 2005 09:36:11 -0800

Louise,

Care should be taken to include any and all necessary autoregressive
structure reflecting dependence on previous values in addition to
deterministic structure reflecting day-of-the-week; hour of the day;
month of the year effects. In additition special events can change
demand not only in a synchronous way but wil both lag and lead effects.
Furthermore there maybe the need to incorporate level shifts, local
time trends , seasonal pulses that were not known apriori reflecting
omitted variables. This approach should also detect unusual values,
such as 1,9,1,9,1,9,5,9 where the 5 represents an "outlier" or more
correctly an "inlier" .

Our software unlike any other routinely provides the answer by
examining the data and constructing minimalyy sufficient models with or
without user guidance.

Please feel free to visit http://www.autobox.com , read up on time
series at http://www.autobox.com/AFSUniversity/afsuFrameset.htm and
download a free copy of FreeFore
at http://www.autobox.com/Freefore-over.htm

You can read some case studies at
http://www.autobox.com/casestudies.htm

If I can help in particular please feel free to call me at 215-675-0652
in theUS

Dave Reilly
Automatic Forecasting Systems