I want to generate a new variable standard error of forecasting (stdf) after VAR and/or SVAR. I can easily generate stdp (standard error of prediction) and stdf after regress, and stdp after VAR. What about stdf after var? Could any one of you please help me?
Hint this is also called standard error of prediction because a forecast is a prediction..
David Booth
Dear Prof. David Booth, thank you so much for your response. However, since there are two separate postestimation commands in Stata - 'stdp' (standard error of prediction) and 'stdf' (standard error of forecasting), it seems to me that stdp should not be identical to stdf. For instance, please follow this link: https://www.stata.com/manuals/p_predict.pdf
How do you plan to use stdf? I would recommend out-of-sample MAE or MSE. Let me recommend this chapter providing guidelines on processing forecasting errors: Chapter Forecast Evaluation Techniques for I4.0 Systems
In stata:
stdp = the standard error of the predicted expected value
stdf = the standard error of the forecast
The problem with stdp and stdf is that these are only estimates and have no empirical support unless out-of-sample rolling origin forecasts are calculated in accordance with the methods described in the chapter above. If your model is not specified 100% correctly (which happens almost always because assumptions are inevitably based on simplifications) then stdp and stdf become non-indicative of real forecasting performance. More on the topic: Chapter Forecast Error Measures: Critical Review and Practical Recommendations
Check https://www.stata.com/manuals/tsvarpostestimation.pdf