VARForecast (VAR forecasting)

[EViews Gareth]

This thread is about the VARForecast Add-in that provides an easy way to perform forecasts from VAR objects. Simulated forecast standard errors are also provided.

[oek_stat]

Hello,
I have got a problem with the forecast. I estimate my var model and then I start the forecast and get the error message "y already exists".
y is the endogeneous variable I want to predict. It exists because I need y for the estimation. What is my fault?

[EViews Gareth]

Could you describe exactly what you're doing?

[terrya]

Hi
I just ran this add-in for a 7-variable VAR for the first time. It doesn't appear to give the standard errors as implied by the documentation.
The series Y_f_se appears for each Y but have only the forecasts for those Ys. I used the command var01.fcast(g,e)_f_se.

[EViews Gareth]

I'm not sure I can replicate the problem. If I run the following code, it appears to work ok:

Code: Select all

rndseed 1
create u 100
series y1=nrnd
series y2=nrnd
series y3=nrnd

var v1.ls 1 3 y1 y2 y3

v1.fcast(g,e) _f _se
show y1_f y1_se


[terrya]

I'm not sure I can replicate the problem. If I run the following code, it appears to work ok:

Code: Select all

rndseed 1
create u 100
series y1=nrnd
series y2=nrnd
series y3=nrnd

var v1.ls 1 3 y1 y2 y3

v1.fcast(g,e) _f _se
show y1_f y1_se


Thanks. I'll try it and see what happens.

[terrya]

Hi.
Just tried it. Message y_f not defined.

[terrya]

Hi
I've found the problem (in part). I didn't notice the spacings after ....(g). This appears to be a problem in the documentation as the command doesn't appear to have spacings. It was you posting that showed it to me when I looked at it carefully. I still get the error message even though it showed the output.

Thanks

[terrya]

Me again. The problem was me not your program.

[ratcliff]

Thanks for the work on this add-in! It's better than the program I wrote to generate the Forecast Evaluation Table for a VAR.

I have one question/suggestion. Say I'm forecasting RGDP from two different VARs, where I select the forecast suffixes as _t and _d. The display names for RGDP_t and RGDP_d both default to something like RGDP(vscen), so legends in graphs with the two forecasts have two lines with the same label. Is there a way to leave the Display Names on the forecasts unassigned? It's easy enough to fix this by hand/with code, but I thought I'd make the suggestion since it's probably easy to fix.

[EViews Gareth]

Hadn't even occurred to me what would happen to the display names - I'll make it something better. Any suggestions on what it should be?

[ratcliff]

Hadn't even occurred to me what would happen to the display names - I'll make it something better. Any suggestions on what it should be?

I'd say just leave the display name blank, so that the labels in the graph just use the name of the series. That works for both the single forecast default case (rgdp_f), as well as the multiple forecasts with different suffixes (rgdp_t vs rgdp_d).

[EViews Gareth]

Actually after I looked at the code, that's exactly what I'm doing if you ask the Add-in to produce graphs.

I'll change it so that it does that whether you ask for graphs or not. Until then, you can always just ask for graph views, and it will handle the displayname appropriately.

[mmezaduering]

Dear Sir or Madam: I'm forecasting with VAR and my question is the following, I have a two VAR model (variabels Y and X) from which I want to forecast, I want to see how a 5 percent increase in X affects Y for the next 4 quarters? I already solve a model but for that purpose there is no need for let's say an exogenous variable the f5 percent increase. How can I simmulate what happens to Y when X changes?

regards

Mynor

[statcon]

Hello,

i am a bit confused about the confidence intervals displayed in the graphical output of this addin. I have run an example and the range of the confidence bands was constant over the total range of observations. My expectation for any kind of confidence/prediction/....-band would be to become wider the farer i move from the middle of the data range.

I would be glad if you could clarify this for me.

Here is one simple example:

Code: Select all

wfcreate test a 1990 2010

rndseed 1
series y1 = nrnd
series y2 = nrnd
series y3 = nrnd

var myvar.ls 1 2 y1 y2 y3


pagestruct(end = @last + 10) *

myvar.fcast(g,e)  _f _se


In this example: Should not the lower and the upper bound move away from each other, exspecially for predictions of future values?