Page 1 of 1
Forecast standard errors are currently only available for non-Bayesian VARs, and are computed via simulation.
Posted: Tue Mar 31, 2026 11:00 pm
by akoh
Hi community, does anyone have any idea how i can calculate forecast standard errors please for a bayesian var? thanks
Re: Forecast standard errors are currently only available for non-Bayesian VARs, and are computed via simulation.
Posted: Wed Apr 01, 2026 3:31 pm
by EViews Kai
Hi,
You can output posterior draws for the forecast into a new page and compute any posterior quantities of interest there.
In your case, to get posterior standard deviation on BVAR forecasts:
1. In the forecast dialog, enter a name for the new page containing your draws in the edit field under "Store each simulation in page (optional)".
2. Go to your workfile and switch over to the page containing your draws.
3. Open the forecast series of interest and go to View > Descriptive Stats & Tests > Stats by Classification....
4. Enter "dateid" in the edit field under "Series /Group for classify" and uncheck both checkboxes inside the "Group into bins if" group.
5. Make sure that the checkbox next to "Std. Dev." is checked. You may also want to check the box next to "Median" if you're using the posterior median as your point forecast so that you can see point estimates next to your standard deviations.
6. Click OK.
Hope this helps.
Best,
Kai
Re: Forecast standard errors are currently only available for non-Bayesian VARs, and are computed via simulation.
Posted: Mon Apr 06, 2026 11:03 pm
by akoh
Hi Kai,
Thanks for your help! After i have carried out the steps and gotten the stats table for mean, median and std deviation, how do i save those 3 columns as new series please? i wish to do some calculations on those new series
Thanks again!
Alfred
Re: Forecast standard errors are currently only available for non-Bayesian VARs, and are computed via simulation.
Posted: Tue Apr 07, 2026 3:46 pm
by EViews Kai
Hi Alfred,
Here are the revised steps. Note that step 4 is different from my previous post. If you have to go through these steps multiple times, you may want to write a program (or programs) to automate as much of this process as you can. Please check the command capture window for guidance.
1. In the forecast dialog, enter a name for the new page containing your draws in the edit field under "Store each simulation in page (optional)".
2. Go to your workfile and switch over to the page containing your draws.
3. Open the forecast series of interest and go to View > Descriptive Stats & Tests > Stats by Classification....
4. Enter "dateid" in the edit field under "Series / Group for classify" and uncheck both checkboxes inside the "Group into bins if" group. Uncheck the checkboxes under "Table Display" in the "Output layout" group.
5. Make sure that the checkbox next to "Std. Dev." is checked. You may also want to check the box next to "Median" if you're using the posterior median as your point forecast so that you can see point estimates next to your standard deviations.
6. Click OK.
7. Select everything within the top and the bottom borders of the table body, including the header row.
8. Right-click inside the series object window and select "Save table to disk...".
9. Set "file name/path". Set "Range" to "Selection" inside the "Options" group. Set "Number format" to "Full precision" in the "Numbers" group. Click OK.
10. In your workfile, switch back to the original page.
11. Drag and drop the saved table file into your workfile. You can either immediately click on Finish or step through the dialog and make changes as you see fit, e.g., renaming the series being imported.
Best,
Kai
Re: Forecast standard errors are currently only available for non-Bayesian VARs, and are computed via simulation.
Posted: Thu Apr 09, 2026 11:45 pm
by akoh
Thanks again! Much appreciated.