[Ulrich Glogowsky]

Hello community.

My answer is not specifically related to the pvar2 packages, but it might interest you. Recently, we have uploaded a Stata module to the ssc-Archive that computes panel vector autoregressions.

The module XTVAR estimates a panel vector autoregression, using a least squares dummy variable estimator. The estimator fits a multivariate panel regression of each dependent variable on lags of itself and on lags of all the other dependent variables. The program also produces Forecast Error Variance Decompositions and Impulse Response Functions. For inference, bootstrap and Monte-Carlo methods are implemented.

Please find the module here: http://econpapers.repec.org/software...de/S457944.htm

Best.
Ulrich

[sarry]

Hi I´m trying to get mediated moderation. In my case my outcome var. is censured in left and right.
Mediator var. is binary and moderador continuos variable. I try to check with this commond: sureg (mv1 iv)(mv2 iv)(dv mv1 mv2 mv3 iv). Is a correct way o no?

[samialameen]

Ulrich,

The <xtvar> program keeps asking for varlist no matter what data set I use. Can you try to run it on the (grunfeld) data set that comes standard with Stata, and check if it works

All the best

[Nick Cox]

"samialameen":

On the evidence Ulrich Glogowsky is only an occasional visitor here. If your question is for him, do send him an email directly.

If you expect others in the forum to be interested, please read the FAQ Advice carefully, especially sections 12 and 6, and start a new thread, as your question is no longer about pvar.

[Leo Caravaggio]

Tim Grünebaum: para poder correr correctamente el pvar es necesario que al menos una de las series sea estacionaria. Si todas son "no estacionarias" habría que tomar la primera diferencia de alguna de ellas. La transformada de Helmert no hace I(0) la serie.

[Marcos Almeida]

Since texts in this forum are supposed to be written in English, and in order to make the last message understood to a broad audience, I decided to (somewhat) translate Leo Caravaggio's message to Tim Grunebaum:

"In order to correctly run pvar it is necessary to have at least one of the series as stationary. If all <series> are 'non stationary' it would be needed to take the first difference from one of them. The Helmert transformation doesn't make the serie to become l(0)".

[Tim Grünebaum]

Thanks to Leo for the advice, thanks to Marcos for translating!

Is it really that simple? I suppose taking first differences eliminates the fixed effects what is not what we intend by using a pvar is it?
After taking 1st diff. we do not need to use helmert transformation as the series is already demeaned.
After doing so we do not need Panel techniques at all because we can pool everything as the FE are gone I might think.
And why taking 1st diff. of only ONE series if all are I(1) ?

This is a farely difficult topic to me somehow ;-)
In the end I see myself taking 1st diff of everything and pooling the thing...

[Leo Caravaggio]

Gracias Marco por traducir, la verdad es que me siento más cómodo escribiendo en castellano, aunque (creo) que no tengo problema en leer en inglés.

Tim, no soy un experto en pvar. Yo también estoy aprendiendo.

Entiendo que funciona igual que un VAR común, necesitás que una de las series sea estacionaria. Por ejemplo una de las series que estoy usando yo es el PIB,
entonces lo supuse I(1) (lo cuál es discutible) y estoy trabajando con la tasa. La otra variable es felicidad, que es I(1), pero la tomo en nivel. Entiendo que lo que estoy haciendo está bien.
Estoy usando un VAR porque lo que quiero ver es el sentido de causalidad de Granger, y necesito el panel por una cuestión de la cantidad de datos (vienen por país, podría promediarlos pero pierdo muchos grados de libertad, son apenas unos 15 países y 15 años).

En todo caso no perdés nada haciendole caso a Inessa y corriendo la transformada de Helmert. Puede ser que no sea necesario (esto ya fue discutido bastante en el foro) pero de todas formas no tiene que tener efectos en los resultados.

Gracias!

[khaled thabet]

Hello all
I used the pvar command to estimate a VAR model for unbalanced panel data and
the command works well. I also seek to identify the causal relationship between productivity and exporting . I used the command pvargranger but I still get the following error message:unrecognized command: pvargranger. so can any one help me! Thnks

[pinar erdem]

Hello All, I couldn't find codes on model selection and model stablity as well as the code for panel granger causality (pvarsoc, pvarstable, pvargranger). I appreciate if someone tells me where I can find ado files for these codes. Thanks!

[Leo Caravaggio]

khaled thabet, pinar erdem, you can check here:
http://econ.worldbank.org/WBSITE/EXT...469382,00.html
https://sites.google.com/a/hawaii.ed...love/home/pvar
and install all the .ado files
any q just ask.
l.

[saeede haji]

hi, can you help me to draw graph for variance decomposition in pvar
i use: pvarfevd, porder( A B)/ for variance decomposition but this command give me table and i need graph for variance decomposition
thank you ery much