Proc Forecast MAPE Calculation

[dinesh_ltjd2]

Hi All,

I have been new with forecasting procedures in SAS. I have been forecasting my data based on 60 data points that I have from Jan'11 to Dec'15 (Monthly data).

I am using "outfitstats" option to calculate the stats for my results. Now this option calculates MAPE based on residuals of all 60 data points nowever I just want to calculate MAPE using just last 6 or 12 Datapoint . Is there a way to limit the datapoints for these stats calculation

Thanks

Dinesh

[sbxkoenk · Accepted solution]

Hi Dinesh,

Are you using PROC FORECAST as the title suggests?

I wouldn't use PROC FORECAST. The FORECAST procedure is obsolete and has been superseded by newer SAS/ETS procedures.

See this link for the alternatives:

SAS/ETS(R) 14.1 User's Guide
Overview: FORECAST Procedure

http://support.sas.com/documentation/cdl/en/etsug/68148/HTML/default/viewer.htm#etsug_forecast_overv...

This said ... the SAS/ETS forecasting procedures and the SAS Forecast Server procedures (PROC HPFxxxxx) can output goodness-of-fit statistics for the FIT region (range/period), for the HOLD-OUT region (if any) and for the FORECAST region (which happens only if the BACK= option is greater than zero).

If you want MAPE (Mean Absolute Percent Error) for only a subset of one of these regions, then you will have to program it. Very simple for MAPE. See Statistics of Fit in SAS/ETS(R) 14.1 User's Guide:

http://support.sas.com/documentation/cdl/en/etsug/68148/HTML/default/viewer.htm#etsug_tffordet_sect0...

In the Time Series Forecasting Menu System (TSFS), you can choose what you want though. It's called an 'evaluation range' in that application. TSFS is part of SAS/ETS.

Evaluation Range

... is the current period of evaluation setting. This is the range of data that will be used to calculate the goodness-of-fit statistics for models fit to the series.

See Part IV. Time Series Forecasting System in the SAS/ETS(R) 14.1 User's Guide

http://support.sas.com/documentation/cdl/en/etsug/68148/HTML/default/viewer.htm#pt04.htm

Kind regards,

Koen