BookmarkSubscribeRSS Feed
T_Rey
Calcite | Level 5

What is the best way to put a PHASEDOWN event in SAS FS?  We know it phased down to 0 and we know when it is supposed to go to zero.  Obvisouly this will be in the forecast horizon, not in the current date time frame.

4 REPLIES 4
TammyJackson
SAS Employee

I assume that you need an event with a downward slope, and you need that event to be 0 past a certain point. Can you be a bit more specific? Is it 0 before a certain point? Is the slope linear or some other function?

TammyJackson
SAS Employee

Is this the effect you wish to model?

 

 

Decaying Ramp

 

 

 

 

T_Rey
Calcite | Level 5

Yes, using SAS FS and only in the forecast horizon

TammyJackson
SAS Employee

If the dummy variable is 1 (constant) for all the observations in the entire historical period, that can* model a level (mean of the series), and that parameter estimate can be used to create a down ramp in the forecast. If the entire down ramp is in the forecast, then you are relying on your knowledge and experience to predict when the ramp will start and end and also the shape of the ramp. That is, there is no history to validate that any portion of the ramp is the correct model. However, if you have many other similar series, you might be able to use those as proxies to get an ideal of the shape of the ramp.

 

*The effect modeled by a dummy variable depends on whether or not the dummy is transformed (i.e. log) or differenced and whether the series is transformed or differenced. The effect should be appropriate for the type of model, additive or multiplicative.  Also, the dummy can be collinear with other effects in the model such as a mean term.

sas-innovate-white.png

Our biggest data and AI event of the year.

Don’t miss the livestream kicking off May 7. It’s free. It’s easy. And it’s the best seat in the house.

Join us virtually with our complimentary SAS Innovate Digital Pass. Watch live or on-demand in multiple languages, with translations available to help you get the most out of every session.

 

Register now!

Discussion stats
  • 4 replies
  • 1617 views
  • 0 likes
  • 2 in conversation