topic Re: predicton in SAS Procedures

predicton

deleted_user — Sat, 29 Nov 2008 09:07:36 GMT

hi all,
I need some help regarding the forecasting the sales.
my working environment is windows2000, sas/base, sas/stat, sas/graph and enterprise guide4.1.

my data like this:
month 2004 2005 2006 2007 2008 2009
APR 94 592 244
MAY 418 627 820 168 407
JUN 916 1050 1134 800 827
JUL 1628 1785 1826 1409 1402
AUG 2326 2950 2661 2873 3464
SEP 8684 8555 9791 8147 7985
OCT 12919 12376 13802 12681 10665
NOV 20760 17523 20791 20372
DEC 30567 24782 28980 28003

i would like to calculate the predicted sales for the 2008 NOV,DEC and 2009 .

could you please recommend the ways to do this.

advanced
Thanks & Regards,
suresh

Re: predicton

Cynthia_sas — Sat, 29 Nov 2008 15:38:19 GMT

Hi:
You may want to read the doc on how to choose the right procedure or task for forecasting. This sample program, for example,
http://support.sas.com/rnd/app/examples/ets/overlay/sas.htm

Compares PROC FORECAST, PROC ARIMA and PROC REG methods of forecasting time series plots.

Depending on how the forecasting tasks work in EG, you may find that your "wide" data needs to be rearranged
from this: [pre]
month 2004 2005 2006 2007 2008 2009
APR 94 592 244 [/pre]
to this:
[pre]
sldate sales
apr2004 94
apr2005 592
apr2006 244
[/pre]

Instead of one observation composed of multiple sales/year values, you may need to have one observation for every date and sale. You can accomplish this when you read the data or using PROC TRANSPOSE.

I recommend starting with the general forecasting concepts that you can find in the documentation or by looking for some user papers on forecasting:
http://support.sas.com/documentation/cdl/en/etsug/60372/HTML/default/tsdata_toc.htm
http://support.sas.com/documentation/cdl/en/etsug/60372/HTML/default/etsug_tsdata_sect021.htm
http://www2.sas.com/proceedings/sugi31/109-31.pdf
http://www2.sas.com/proceedings/sugi28/009-28.pdf

cynthia

Re: predicton

deleted_user — Sun, 30 Nov 2008 10:31:22 GMT

Thanks a lot for your help.

Re: predicton

deleted_user — Sun, 30 Nov 2008 11:57:33 GMT

Hi suresh,

its better to try on proc Frocast and the best one is proc regression and use prediction VARIABLE what ever u want and use proc T-test for comparision

chandra alapati

Re: predicton

deleted_user — Sun, 30 Nov 2008 17:39:27 GMT

hi,

i tried the methods recommended y you.
i trailed by taking the values of 2004,2005 and 2006.
predicted for the 2007
i got the values like this
Obs _TYPE_ _LEAD_ mon1
1 FORECAST 1 26035.03
2 FORECAST 2 23988.53
3 FORECAST 3 22599.23
4 FORECAST 4 21690.65
5 FORECAST 5 21133.71
6 FORECAST 6 20833.98
7 FORECAST 7 20722.38
8 FORECAST 8 20748.41

these values are so far from the 2007 values.

Could you please comment or recommend any suggestions to do my task well.

Thanks