Elder-Serletis(2010) VAR-GARCH-M

[TomDoan]

If you're interested in looking at what happens when you go to higher order multivariate GARCH, use the GARCHMV.RPF example and change the p's and q's in those examples. Elder-Serletis is already very complicated and needs restrictions to work properly as it is; there is no chance that it will work if the GARCH model is made even more complicated.

[lhydyz]

Dear Tom,
I am new to rats and I am working with the VAR-GARCH-M model promoted by Elder and Serletis. The problem is that when I use trial version 9.2 I can run the programme but when I use the pro 8.0,there is some error. The error is shown below:
## SX11. Identifier %MODELLAGMATRIX is Not Recognizable. Incorrect Option Field or Parameter Order?
>>>>i)=%modellagmatrix(<<<<
I whould appreciate it if you could give me feed back soon~

[TomDoan]

%MODELLAGMATRIX isn't in version 8. If you plan you use that as a permanent version, you should have it updated.

[lali62]

Hi,
Could you tell me how to generate the data text file. As mentioned in the paper, they use logarithmic first differences of the real price of oil and real GDP. Checking the GDP numbers, they do not seem to match
with the text file.

[TomDoan]

Hi,
Could you tell me how to generate the data text file. As mentioned in the paper, they use logarithmic first differences of the real price of oil and real GDP. Checking the GDP numbers, they do not seem to match
with the text file.

That's the data file that we got from the authors. Historical values of GDP growth rates change, sometimes quite a bit, with revisions and benchmarking, so unless you are able to come up with the exact issue of GDP that they used, you won't be able to re-generate the data.

[lali62]

Hi,
Could you tell me how to generate the data text file. As mentioned in the paper, they use logarithmic first differences of the real price of oil and real GDP. Checking the GDP numbers, they do not seem to match
with the text file.

That's the data file that we got from the authors. Historical values of GDP growth rates change, sometimes quite a bit, with revisions and benchmarking, so unless you are able to come up with the exact issue of GDP that they used, you won't be able to re-generate the data.

That I understand, but if I look the magnitude, for e.g let's say for Q3-2007, it shows a value of 4.79. I believe they have used q/q annualized rates instead of log-differences

[TomDoan]

Probably 400*log(x/x{1})

[lali62]

Probably 400*log(x/x{1})

Thanks for explaining this. I have another doubt, is there a way to extract the data in the plots as csv or text file.

[TomDoan]

Use @MCPROCESSIRF rather than @MCGRAPHIRF.

[grkn7878]

Hello,
I have 4 variables. I would like to estimate a model like this
frml bfrml = ||$
1.0,0.0,0.0,0.0|$
b21,1.0,0.0,0.0|$
b31,b32,1.0,0.0|$
b41,b42,b43,1.0||
first one is oil price.
I would like to extend IRF graph with an error band for 4 variables. How can I extend Random Walk Metropolis codes.

Warm Regards.

[TomDoan]

That's a long way short of being "the model". As pointed out in the initial post, there's quite a bit about their model which is specific to it being a two variable model with one of those variables having no discernible GARCH properties (GDP). The Cholesky factor GARCH model probably works OK in that case, but, in general, SVAR-GARCH models don't work as well as more standard forms.

[istiak]

Dear Tom,
When I get Graph 2 (response to positive and negative shock with and without M terms), there is no confidence interval associated with the graphs. How can I show a confidence interval around both responses (with and without M terms)?
Thank you.

[TomDoan]

That's not the point. There's a separate process in that program for doing the full model with confidence bands. The sole purpose of that graph is to demonstrate that the inclusion of the M effects changes responses to no longer be symmetrical between positive and negative shocks.