Forecasting using SAS Forecast Server, SAS/ETS, and more

How to interpret VAR output

Reply
Frequent Contributor
Posts: 98

How to interpret VAR output

Hi all,

I'm running a vector autoregressive (VAR) model) to use digital marketing activities to predict sales. As this is my first time using SAS for VAR, I have some difficulty in interpreting the output.

First, I used the following syntax to check length of lags:

proc varmax data=sales;                                                                                                      

  id week interval=week;                                                                                                      

  model lsales lemails/p=3 noint lagmax=9                                                                                          

                        print=(parcoef pcorr pcancorr);                                                                            

  output out=forecast lead=4;                                                                                                          

run;

The partial correlation output shows AR orders of p=1, p=4 and p=9 are significant. Which order should I go with?

Then I used the following syntax to see the results of VAR(1), VAR(4) and VAR(9):

proc varmax data=sales;                                                                                                      

  id week interval=week;                                                                                                      

  model lsales lemails/p=1;                                                                                                      

  output out=forecast lead=4;                                                                                                          

run;

The estimates on "lemails" vary dramatically in the three models.

Could anyone give me some suggestions?

Below is some sample data:

weeklsaleslemails
19.51169.7189
29.730710.0534
39.820310.0726
49.888410.0837
510.021710.0429
69.699810.1137
79.985810.1397
810.143410.0665
910.359910.1023
109.789810.1726
119.985210.0869
1210.056510.0856
1310.369510.0597
149.878610.0593
159.971710.0242
1610.057510.0198
1710.202310.0751
1810.14629.9976
199.87909.9551
2010.00419.9881

Thank you!

Ask a Question
Discussion stats
  • 0 replies
  • 581 views
  • 0 likes
  • 1 in conversation