Page 4 of 5

Posted: Wed Feb 14, 2018 12:55 pm
hey dakila,

I could solve the problem on my own. Apperently this only appears when using eviews 9. In eviews 10 everything works perfectly fine. Another very important question is how you handled the forecast error variance decomposition. I know how it works in standard VAR models but I never did that in a FAVAR since the variables arent directly included in the VAR. Am I right to suppose that you cant just multiply the standard Forecast error variance decomposition by the factor loadings? In the BBE paper the formular is quit confusing, since its seem like this is exactly what they did. How did you handle that

Best,
Ben

Posted: Thu Feb 15, 2018 12:23 am

Code: Select all

`vardecx=lam.@t*varf*lam          ' variance in x's due to monetary shockvardecxtot=lam.@t*varftot*lam    ' variance in x's due to all commom factors                vardecxcom=@ediv(@getmaindiagonal(vardecx), @getmaindiagonal(vardecxtot))`

Posted: Fri Feb 16, 2018 3:03 am
Thanks for sharing parts of your code. So basically you calculated the part the forecast variance of the variable explained by the monetary policy shock relativ to the variance explained by all factors and the monetary policy shock? Thanks for your help and providing an insight into your code

Ben

Posted: Fri Feb 16, 2018 5:30 am
Hey its me again,

we are struggling with heteroskedasticity in our favar and we know that there are bootstrap meausures which account for that when constructing confidence bands. Could you tell/show us, which bootsstrap measure you applied?

Posted: Sun Feb 18, 2018 3:36 pm
see Killian (1998).

Posted: Mon Apr 09, 2018 10:41 am
Hello,
is it possible to obtain the IRF for a variable which is only differenced and not log-differenced (has negative values) using your add-in, please?
My second question is connected with the IRF interpretation. Is the y axis in standard deviations of original variables or in standard deviations of variables trasformed to exhibit stationarity (for example: if the variable is log-differenced, is the y-axis unit of IRF standard deviation of level or standard deviation of log of the original variable)? Thank you very much.

Posted: Wed Apr 11, 2018 8:39 pm
Hi,
It now includes the first difference option (code=2).Please update the favar add-in.
Y axis is in standard deviation unit (0 mean, 1 variance) of untransformed variable.

Posted: Fri Apr 13, 2018 11:38 am
Thank you very much for your reply. Unfortunately, I can't update the add-in anymore as it says: "Could not open file due to: unknown error." I tried both Eviews 9 and 10 at work. At the same time, I can update other Eviews add-ins without any issues, so it seems that this error is specific to FAVAR addin only at the moment.

Posted: Fri Apr 13, 2018 11:47 am
Try now.

Posted: Fri Apr 13, 2018 11:53 am
It works now. Thank you very much!

Posted: Wed Nov 28, 2018 12:07 am
Hello;

Is it possible to see factor loadings or where can be found? And is it possible to estimate VAR with dummy variables or exogenous variables?

Thanks..

Posted: Wed Nov 28, 2018 8:05 am
You can do this manually

Posted: Thu Nov 29, 2018 12:05 am
Hi Dakila;

Thank you for the quick response. But where exactly can I find the loadings? I try to classify the factors but I did not find that which factor is affected by which variable. Thanks.

`equation eq.ls series01 _facrot1 _facrot2 _facrot3 ffr`