Forecast from VAR

[joaopaulocaetano]

Good morning everyone. I would like to get you a great help for the realization of a pet. I'm working with vector autoregressive model (VAR) with all procedures performed. My question to this forum is: Can from the built VAR model to estimate future values for miha variable of interest? ie, I, from the constructed model, making future scenarios for my study variable. In this sense, I would like someone to help me to perform this procedure in eviews. I thank you for the cooperation of all.

[EViews Gareth]

Use Proc->Make Model from the VAR, then solve the model over the period you wish to forecast over.

[joaopaulocaetano]

EViews Gareth, Thank you for your help. Now see if you can clarify a methodological issue for me. Imagine that the series of my model are quarterly and will, for example, from 1st quarter 2000 to 4th trim 2014. And made the model, I want to design to the 4th quarter of 2015; my question is this: I need to have any prior information to the period I want to project or all of my series can stand up to the 4th trim 2014 and when I run the model they are designed to trim 4th 2015?

[trubador]

I do not follow the question, but it sounds like the answer lies within the domain of dynamic vs. static forecasting. If so, please read through the discussion that starts with this post: http://forums.eviews.com/viewtopic.php? ... 219#p41024

[joaopaulocaetano]

I had already looked at the indicated post. The problem is that there we work with models of ARIMA type - adapted to the GARCH version -. In this case, the forecasting process is simple and straightforward in eviews. In my case, I want to forecast values from a VAR (vector autoregressive) which has no direct explanation in eviews to be performed. To do so, must first solve the system and then indicate the variables that I want to forecast (in this case I'm working with dynamic intertemporal model, hence the difficulty). In other words, I am working with endogenous and exogenous variables, which has left me confused as to the correct parameters to perform the projection of my values.

[trubador]

Actually, the gist of forecasting approach remains the same. The model object offers a more general platform to carry out forecasting and simulation. So the discussion holds: If your exogenous variables do not have future values for the forecasting period then you cannot perform the forecast at all. Since VAR (or ARIMA) models are dynamic by design, they have the ability to generate future values for the endogenous variables through exploiting the estimated/solved parameters of the system.

[joaopaulocaetano]

That'S Perfect. Now I understood the mechanism. Thank you very much.