topic Re: Forecast Technique in SAS Forecasting and Econometrics

Forecast Technique

parmis — Thu, 14 Feb 2019 16:30:35 GMT

Hello,

I'm trying to forecast the number of X and Y for each ID monthly for the next two years.

I have 6 months of data, my numbers are pretty much flat.

I'm not sure what forecast technique I should be using.

Thanks for your help in advance.

Date	ID	X	Y
Jan-18	A1	10	4
Jan-18	B1	27	5
Jan-18	C1	35	22
Jan-18	D1	34	27
Jan-18	F1	24	27
Feb-18	A1	2	44
Feb-18	B1	33	20
Feb-18	C1	0	7
Feb-18	D1	24	44
Feb-18	F1	0	34
Mar-18	A1	42	24
Mar-18	B1	67	2
Mar-18	C1	6	33
Mar-18	D1	0	0
Mar-18	F1	63	24
Mar-18	L1	36	0

Re: Forecast Technique

imvash — Thu, 14 Feb 2019 17:45:17 GMT

Since you do not have a long history of data, I suggest to calculate forecast values using a simple method like moving average. You can use PROC ARIMA with the options: noint noest and method=CLS. See the documentation at:

http://support.sas.com/documentation/cdl/en/etsug/60372/HTML/default/viewer.htm#etsug_arima_sect017.htm

Assuming you have at least 6 months of data (since proc ARIMA needs at least 6 observations to perform identification step), you may use the following code:

data test;
   input date monyy7. ID$ X Y;
   format date monyy7.;
   datalines;
JAN2018 A1  10  4
JAN2018 B1  27  5
JAN2018 C1  35  22
JAN2018 D1  34  27
JAN2018 F1  24  27
FEB2018 A1  2   44
FEB2018 B1  33  20
FEB2018 C1  0   7
FEB2018 D1  24  44
FEB2018 F1  0   34
MAR2018 A1  42  24
MAR2018 B1  67  2
MAR2018 C1  6   33
MAR2018 D1  0   0
MAR2018 F1  63  24
APR2018 A1  35  24
APR2018 B1  22  2
APR2018 C1  14   33
APR2018 D1  20  0
APR2018 F1  13  24
MAY2018 A1  21  24
MAY2018 B1  70  2
MAY2018 C1  60  33
MAY2018 D1  55   0
MAY2018 F1  55  24
JUN2018 A1  41  24
JUN2018 B1  64  2
JUN2018 C1  60  33
JUN2018 D1  10  0
JUN2018 F1  33  24
;

proc sort data = test out = test;
	by ID;
run;

proc arima data = test plots = none out = outX;
	by ID;
	identify var = X;
	estimate p = (1 2 3) ar = (0.3333 0.3333 0.3333) noint noest nostable method = CLS;
	forecast lead = 24;
run;
quit;

proc arima data = test plots = none out = outY;
	by ID;
	identify var = Y;
	estimate p = (1 2 3) ar = (0.3333 0.3333 0.3333) noint noest nostable method = CLS;
	forecast lead = 24;
run;
quit;

Re: Forecast Technique

parmis — Thu, 14 Feb 2019 18:23:57 GMT

Thanks for your reply. Do I have to write a separate code for each variable(X and Y)?

Also, when I run the query I get the following error: Forecasting was not performed because estimation was not done

Re: Forecast Technique

imvash — Thu, 14 Feb 2019 18:27:22 GMT

Did you run my code and received the error?

As far as I can remember, you cannot specify several variables in proc arima. You need to do it separately for each variable.

Re: Forecast Technique

parmis — Thu, 14 Feb 2019 18:28:42 GMT

yes , I didn't make any changes

Re: Forecast Technique

imvash — Thu, 14 Feb 2019 18:47:13 GMT

I don't know why it doesn't work for you. I have it running on both SAS Studio and SAS 9.4.

Re: Forecast Technique

ballardw — Thu, 14 Feb 2019 20:23:19 GMT

@parmis wrote:

Thanks for your reply. Do I have to write a separate code for each variable(X and Y)?

Also, when I run the query I get the following error: Forecasting was not performed because estimation was not done

Go to the log, copy the log entry from the first data step using your test data set code through the end. Paste it into a code box opened with the {I} forum icon.

If you do not get the same message then there is something different in the data used.