Hi,
I am developing an international tourism demand forecasting model using ARIMA/SARIMA technique using STATA and Eviews softwares. However, in both softwares, forecast evaluation is a bit tricky. Is anyone aware of how to evaluate the forecasting accuracy/performance using STATA and Eviews. I am interested in using the MAPE, RMSE etc to evaluate the forecast accuracy.
In my short experience:
rmse: Low values are better, meaning that the gap between Y and YF is smaller.
Theil inequality: If = 0, Y and YF are equal. If = 1, the predictive power is bad.
Mape : if = 15, on average, the forecast is off by 15%. It is used for comparative purposes. It is used to compare the fit of your model with other time series models.
Hopefully, more experienced guys will do futher clarifications.
I remember EViews has that capability but I didn't use it after version 3. I can check in my notes if you don't find it. But Stata (at least version 12) has indeed no standard way to compare forecasts to observations. I had to solve that problem for a course I gave of time series with Stata some years ago and I found a Robert Yaffee contribution, see https://ideas.repec.org/p/boc/usug10/10.html. I used it but it was not so easy.I have a text (also on R) in French about my experiments that I can share with you. Contact me directly.
https://ideas.repec.org/p/boc/usug10/10.html
Forecast evaluation in Eviews: save the forecast series and hence compute a series of forecast errors. Measures such as MAPE, RMSE can then be calculated from the series of errors. RMSE: form a series of squared errors and take the square root of its average. MAPE: form a series of % errors, transform to absolute values and note the average.
You may use in Stata the Diebold Mariano Test.
ssc instal dmariano
and then check it out the help.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series