topic Re: Regarding Forecasting in SAS Forecasting and Econometrics

Regarding Forecasting

Anil_K — Wed, 15 Jul 2009 05:38:01 GMT

Hi,
If i an forecasting in SAS HPF,for the following form of data
monyy demand
jan01 12
mar01 17
jun02 19

then is it necessary to make it contninous ie jan01,feb01,........,and replace missing values with zero or we put as it is and forecast.

Re: Regarding Forecasting

udo_sas — Thu, 16 Jul 2009 07:56:41 GMT

Hello -
You might want to consider to either pre-process your data using PROC TIMESERIES or to use the "accumulate" and "setmissing" options in the HPF procedures.

Example:
data test;
input date date7. demand;
format date date7.;
cards;
01JAN09 10
01FEB09 12
01APR09 12
01MAY09 10
;
run;
proc timeseries data=test out=series;
id date interval=month accumulate=total setmissing=0;
var demand;
run;

or when using PROC HPF for example:
proc hpf data=test ...;
id date interval=month accumulate=total setmissing=0;
...
run;

Regards,
Udo

Re: Regarding Forecasting

Anil_K — Fri, 17 Jul 2009 05:03:45 GMT

Hi,
thanks for ur reply but my question is little bit different.

Suppose when using PROC HPF :
proc hpf data=test ...;
id date interval=month accumulate=total setmissing= ;
...
run;
then it also provide the result.So my question is:
1. is it necessary to set missing zero?
2.if we set missing as blank then it provide the different answer as compared to set missing as zero.

Re: Regarding Forecasting

udo_sas — Mon, 20 Jul 2009 07:10:21 GMT

Hello,
Please excuse for missing the point. You will find lots of information about how missing values are treated in Online Help. I'm copying some relevant sections for your convenience.
Thanks,
Udo

> 1. is it necessary to set missing zero?
From Online Help: "Sometimes missing values should be interpreted as unknown values. The forecasting models used by the HPF procedure can effectively handle missing values. But sometimes missing values are known (such as when missing values are created from accumulation), and no observations should be interpreted as no (zero) value. In the former case, the SETMISSING= option can be used to interpret how missing values are treated. The SETMISSING=0 option should be used when missing observations are to be treated as no (zero) values. In other cases, missing values should be interpreted as global values, such as minimum or maximum values of the accumulated series. The accumulated and interpreted time series is used in subsequent analyses."

> 2.if we set missing as blank then it provide the different answer as compared to
> set missing as zero.
From Online Help: The treatment of missing values varies with the forecasting model. For the smoothing models, missing values after the start of the series are replaced with one-step-ahead predicted values, and the predicted values are applied to the smoothing equations.