

7. UNIVARIATE MODEL FORECASTING › 7.5 Component Operation › 7.5.3 Transform Historical Observations Screen
7.5.3 Transform Historical Observations Screen
If you specify YES for extended options on the Univariate
Model Forecasting screen, you will have the opportunity to
specify the transformation of historical observations. Using
this option, you can specify either GMA smoothing, a LOG10
transformation, or a user-defined transformation. Specify
NONE if you do not want smoothing or transformation. NONE is
the default.
If you specify USER for this option, then you are required to
provide the SAS statements to perform the transformation
and/or the reverse transformation by specifying a Y for at
least one of the last two options on this screen.
/--------------------- Transform Historical Observations ---------------------\
|Command ===> |
| |
|Inquiry Step: Univariate Model Forecasting |
| |
|Smoothing technique ===> NONE (GMA/LOG10/USER/NONE) |
| |
| |
| |
|Geometric moving average data smoothing parameters: |
| Equation alpha value ===> ____ (.01 - .99) |
| Initial value ===> ________ |
| |
| |
| |
|User defined data transformation SAS statements: |
| Statements to transform historical data ===> N (Y/N/R) |
| Stmts. to reverse history transformations ===> N (Y/N/R) |
| |
| |
| |
| |
| |
\------------------------------------------------------------------------------/
Figure 7-12. Transform Historical Observations Screen
The Transform Historical Observations screen allows you to
modify the historical data series prior to the modeling
process. (The processing of historical data observations is
discussed in Section 7.4.2.) You can choose from three
different types of optional processing:
o Data smoothing using a geometric moving average
o A log base 10 transformation of the historical
observations
o A user-specified exit to process the historical
observations and possibly a corresponding exit to reverse
the transformation process after the forecasting process
If you specify that the data is to be smoothed using a
geometric moving average, then one other parameter is
required and one more is optional. You must specify the
value of alpha (0 < alpha < 1) to be used in the geometric
moving average equation (Equation 10). We recommend that you
initially try a value of 0.5.
In addition, you may optionally specify an alternative value
for the initial historical observation. This value is used
to replace the first value when you do not believe that the
first point of the historical data series is representative.
If you do not wish to have an alternative value, leave this
parameter blank.
If you specify that the historical observations are to be
transformed using a LOG10 function, no other parameters are
required or allowed on this screen.
If you specify that the historical observations are to be
transformed using your own SAS code, you need to indicate on
this screen which type of transformation you wish to include:
o Statements to transform historical data
o Statements to reverse the transformation after the
analysis
If you select either of these transformations, the
appropriate screen is displayed to allow you to enter your
own SAS code. These screens are discussed in Section 7.5.4.
Copyright © 2014 CA.
All rights reserved.
 
|
|