What's the difference between all the time series procs?

#1
I'm kind of curious as to specific differences in the all of the many time series procedures in SAS. I know, for instance, there is PROC ARIMA, PROC FORECAST, PROC TIMESERIES, PROC X12/X11.........

I know there is a document (http://support.sas.com/publishing/pu...haps/57275.pdf) that someone one here suggested to me to try and help determine which procedure to use for a specific case, but I am still sort of confused about when it is best to use a certain PROC.

For instance, if forecasting is the goal, does it make sense to just use PROC FORECAST? I know you can add forecast statements to other time series procedures as well which would suggest you could get the same information using multiple procedures.

Is there any type of chart (or something) which perhaps compares and contrasts each of the main SAS procedures for time series (possibly giving examples of when to use one procedure over another)? Just curious.

Thanks!
 

noetsi

Fortran must die
#2
My answer to this, it's a general answer when choosing with SAS, is to find the method you want to use first based on what is considered the best one or the one you are most familiar with and then find which Proc does it. Where a method is in SAS (which Proc) is largely artificial IMHO.
 
#3
So as someone who has never really used SAS to do time series (my work does not own the SAS module to access these PROCs, though I think (hope) that will change once we upgrade to a newer SAS version) you are essentially saying that each of the time series procedures can do the same thing? It's just basically GPLOT vs. SGPLOT - you can pretty much do the same thing with either but they have just slightly different syntax?
 

noetsi

Fortran must die
#4
No they won't neccessarily do the same thing. What I meant is that you need to find out what specifically you want to do first, say ARIMA, then find out where this is located in the proc (it's in proc ARIMA in that case). Different types of time series (ARMA, ARIMA, exponential smoothing etc) have different assumptions, methods and results. The key is to understand the method you want to use first, then simply find where SAS has stuck it.

Before you try to learn the SAS code (which for time series runs hundreds if not thousands of pages) you need to learn time series methods (if you don't know these already in detail) and which you prefer to use. Once you understand them its not overly difficult to find that method in SAS - the documentation covers this.

Remember that SAS, splits its modules up. I had SAS stats, but could not run the time series module (ETS) until we paid for that seperately.
 
Last edited: