Time varying SVAR

[dakila]

This thread is about the tvsvar add-in that estimates time varying structural VARs.

[Nas1]

Hello,

thank you for posting the add-in.
why I am getting the message of "out of memory" error. is this mean that I am having a problems in the matrix? I am using a three randomly generated data; the number of data 200. the number of gibbs sampling iteration 10000 and number of burn-in draws 2000. everything else is as default.

I am using eviews 9 with latest update was in November 2015.

another point, please clarify what is Date selection vector really mean?

Kind regards,

[dakila]

1. Why did you use randomly generated data?. There is an example data in the tvsvar folder if you want to test this add-in. I don't know why you getting this error. I guess this is because of your computer.

2. The tvsvar estimates time varying structural VARs. So you can get impulse response function for any date you want. If you want the impulse responses for 1980q1, 1999q3 and 2007q4, just insert 1980q1 1999q3 2007q4 on the box with no comma.

[Nas1]

Hello,

Thank you so much for answering my inquiry concerning Date selection vector .. the adds in was already uploaded and I did not notice the example in the forum, thats why I used a randomly generated data to check the adds in.

Do you know how to integrate Convergence Diagnostics test and the marginal likelihood estimation in TVP VAR ?


Kind regards,

[dakila]

Sorry, I don't know.

[nadybe]

hello everyone,

I tried the TVSVAR Add in with my data in first-differences and it return a near singular matrix. Do you know how to fix this issue?
Does anyone have informations on preliminary analysis before using the time varying SVAR?

Thank you

[dakila]

try to estimate with data in level or annual growth.

[wjgatt]

Hi, thanks for this great add-in.

Is it possible to save all the results as series, so that they can be exported? The program which runs the add-in is encrypted, so I cannot modify the code to have Eview save the results as series for example.

Thanks
W

[dakila]

Yes. It is possible. I will release version 2.0 soon.

[wjgatt]

Awesome, thank you!

[Roody36]

Hi,

When I run the example on the SVAR - the error "Matrix-Vector index is out of range" occurs - Do you happen to know what might be the reason?

Thanks

[dakila]

Which version do you use? Probably you use the older version. I think you need to upgrade your Eviews to version 9.

[Roody36]

Well you may be right , I use Eviews 8 - The add-in is compatible only with higher version?

[dakila]

Yes

[pascalle]

Hi,

I am using eviews 8 to estimate the TVC-SVAR model with monthly data 2000m1 to 2015m12. There was an error message: "Sizes do not match in matrix function" during estimation. May I know how can I fix this problem? And what does the "y01" represent?
I find many papers using 40 as a training sample. what does training sample mean? Is there any test to measure it?

Thank you.