Page 1 of 4

Large Bayesian VAR

Posted: Mon Nov 28, 2016 3:19 pm
by dakila
This thread is about the lbvar add-in that estimates Large Bayesian VARs.

Re: Large Bayesian VAR

Posted: Wed Aug 30, 2017 3:24 am
by awatt43
Hi there,
I keep getting the following error "Matrix-Vector index is out of range" what does this refer to?
Thanks,
Abi

Re: Large Bayesian VAR

Posted: Wed Aug 30, 2017 6:42 pm
by dakila
if your data have missing values then adjust the sample size.

Bayesian VAR

Posted: Mon Sep 04, 2017 6:53 pm
by adrangi
Hi. I installed the Add-in BVAR. I tried to run it for 4 variables, 6 lags. Nothing happened. Any ideas? Thanks. BA

Re: Large Bayesian VAR

Posted: Mon Sep 04, 2017 7:44 pm
by dakila
This thread is about the LBVAR add-in not BVAR.

Re: Large Bayesian VAR

Posted: Tue Sep 05, 2017 10:57 am
by adrangi
Hi . Sorry, my search landed me here so I thought covers both. I've installed LBVAR. Will try both. Thanks. BA

Re: Large Bayesian VAR

Posted: Tue Sep 19, 2017 2:58 pm
by erhardmenker
Thanks for creating such a useful add-in for my research needs. Do you have an unencrypted version of the LBVAR program that can be viewed? I am trying to view the inner-workings of the code (e.g. open intermediate objects, view portions of the algorithm) for the purpose of understanding the add-in.

Re: Large Bayesian VAR

Posted: Thu Sep 21, 2017 5:22 pm
by dakila
can you check your private message?

Re: Large Bayesian VAR

Posted: Thu Oct 12, 2017 9:32 am
by vikas.trades25
I am using a 10 variable, with 5 lags model with quarterly data from 1980 onwards, All my variables enter in log-differences, so random walk prior is
0. Do I need to enter an array of 10 0's for all the 10 variables in the "Random walk prior box". Lambda is 0.15 and Tau is 1.5. Estimation is forecasting.
Forecast horizon is 1 quarter. I do not have any endogenous variables. What do i need to give in "Fit evaluation variables" box.

I also want to see the fitted values. How do I see that.

Thanks

Re: Large Bayesian VAR

Posted: Thu Oct 12, 2017 1:38 pm
by vikas.trades25
I get the following error "Size do not match in matrix function". 9 variables are in log differences while one is in levels.

Re: Large Bayesian VAR

Posted: Thu Oct 12, 2017 3:48 pm
by dakila
You need to create the vector for Random walk prior box. For example:

Code: Select all

vector irw=@zeros(10)

If fifth variable is non-stationary then

Code: Select all

irw(5)=1

You manually entered lambda and tau. So you don't need to fit evaluation variables. If you choose grid search then you need the fit evaluation variables.
Fit evaluation variables is usually GDP, Price(CPI or GDP deflator) and Interest rate (Fed Fund rate). You can choose whatever variables you want. But don't choose too many variables. Approximately 3 variables are ok.
You want to see the forecasted values? it is the variable with suffix _f.
Please read the instruction document which is located in the C:\Users\...\Documents\EViews Addins\LBVAR folder (lbvar.pdf).

Re: Large Bayesian VAR

Posted: Fri Oct 13, 2017 7:34 am
by vikas.trades25
All my variables are stationary. I have created the "irw" vector. Now do I need to give all the variable names in the endogenous variables, since all the variables are included in the model. It gives me the same error "sizes do not match in matrix function".
When I also include irw in the endogenous variable list, it gives me the error "irw is not a series".

How to solve this. All the 10 variables are to be included in the model.
Also how to see the fitted series. Is there a way to get this.

Re: Large Bayesian VAR

Posted: Fri Oct 13, 2017 3:13 pm
by dakila
You should put the irw vector into the Random walk prior box.

Re: Large Bayesian VAR

Posted: Mon Oct 16, 2017 8:10 am
by vikas.trades25
Thanks for the previous reply.
I get the following error now: "EXTSER is not the same length as sample"
I need to get the forecasts for all 13 vars. Do I need to give anything in the impulse variable box.

Re: Large Bayesian VAR

Posted: Tue Oct 17, 2017 6:30 am
by dakila
If you have the missing observations because of the first order difference, then you should adjust the sample size.