Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# st: Is forecast evaluation with stata possible?

 From tom smith To "statalist@hsphsun2.harvard.edu" Subject st: Is forecast evaluation with stata possible? Date Sat, 1 Dec 2012 00:04:36 +0000

```Hello,

I have a question concerning the following:

I estimated a number of forecasting models using different methods and variables. I did this a large number of times on an every time larger sample. More specifically I created a loop in Stata that estimates the different models on a first sample and then does the forecasting for 4 periods on the basis of that first sample, then the different models are again estimated based on a sample that is 4 periods larger and then again forecasting is done for 4periods etc..

So now I ended up with a number of different variables that contain 10 different forecasts of 4periods long under each other.

Date:  Forecast  Forecast   Actual variable
Model1    Model2

...
y1Oq1     F1       F1           x
y10q2     F1       F1           x
y10q3     F1       F1           x
y10q4     F1       F1           x
y11q1     F2       F2           x
y11q2     F2       F2           x
y11q3    ...      ...          ...
y11q4
y12q1
y12q2
y12q3
y12q4
...
y10q4

Now my question is how I can evaluate these forecasts, I want to see which is the best forecast. I also want to see which forecast is the best for each period.

Does anybody have an idea how I can do this? Normally I could use RMSE, MAE and MAPE but I don't know how I can apply them to my results. And I don't know how to calculate them with Stata.
When I want to evaluate separately the forecast of the 4 forecasted periods and would  calculate the RMSE, MAE and MAPE I also don't know which numbers I have to use for the parameter representing the size of the forecast.

I mean the N here: http://en.wikipedia.org/wiki/Forecasting#Forecasting_accuracy
Should I use "4" the lengt of my forecasts or "10" the total number of forecasts I want to evaluate?

Or is it possible that forecast evaluation is just not possible with stata?

Kind regards,
Tom
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
```