Bai Perron Test

Discussion of models with structural breaks or endogenous switching.

Bai Perron Test

Postby bizi_rats_10 » Fri Aug 06, 2010 11:52 am

Hi All,

I am trying to use the Bai Perron test to check break points in a VECM of 5 variables. However I am not getting the results I am expecting. I have hard time specifying the list of regressors; is this the number of variables. Second how do I get a graph showing break points with this test? If anyone knows how to go about this please let me know I would appreciate that.

Thanks,

JC
bizi_rats_10
 
Posts: 10
Joined: Sun Jul 18, 2010 3:47 pm

Re: Bai Perron Test

Postby TomDoan » Fri Aug 06, 2010 4:05 pm

Bai-Perron is specifically for tests in a (single) linear regression. You could apply it to individual equations in error correction form, but that's only if you assume the cointegrating vectors are fixed, and you're looking to see if the short-run dynamics or the loadings on the error correction terms change.
TomDoan
 
Posts: 2720
Joined: Wed Nov 01, 2006 5:36 pm

Re: Bai Perron Test

Postby bizi_rats_10 » Sat Aug 07, 2010 12:16 am

Thanks very much. I will try it and see what happens.

JC
bizi_rats_10
 
Posts: 10
Joined: Sun Jul 18, 2010 3:47 pm

Re: Bai Perron Test

Postby bizi_rats_10 » Wed Aug 11, 2010 10:56 am

Hi,
Following up on the Bai Perron Test, I did test the break point in each of the equations (individually) of a VAR instead of VECM and I think it worked. My other question is how do you determine the option "minspan"? It says that minspan = shortest distance between two breaks; but how about if you have one break or if you do not know the break point?
Thanks!

JC
bizi_rats_10
 
Posts: 10
Joined: Sun Jul 18, 2010 3:47 pm

Re: Bai Perron Test

Postby TomDoan » Wed Aug 11, 2010 12:32 pm

If you have only one break, minspan only affects how close to the ends of the data set that you are permitting a break. The default is the number of regressors---it can't be shorter than that since you need that many in order to estimate the model between consecutive breaks. Unless you have only one regressor, that usually is fine.
TomDoan
 
Posts: 2720
Joined: Wed Nov 01, 2006 5:36 pm

Re: Bai Perron Test

Postby iloverats » Sun Jul 24, 2011 10:36 am

Dear
Does bai perron test (Baiperron.src)report the critical value for supF test?
iloverats
 
Posts: 45
Joined: Thu Dec 02, 2010 11:32 am

Re: Bai Perron Test

Postby TomDoan » Mon Jul 25, 2011 5:09 pm

iloverats wrote:Dear
Does bai perron test (Baiperron.src)report the critical value for supF test?


No. Bai and Perron have rather large tables for the critical values in their papers.
TomDoan
 
Posts: 2720
Joined: Wed Nov 01, 2006 5:36 pm

Re: Bai Perron Test

Postby iloverats » Thu Aug 11, 2011 12:38 pm

TomDoan wrote:
iloverats wrote:Dear
Does bai perron test (Baiperron.src)report the critical value for supF test?


No. Bai and Perron have rather large tables for the critical values in their papers.


dear
may you tell me how to choose the minspan value in baiperron.src?
thanks
iloverats
 
Posts: 45
Joined: Thu Dec 02, 2010 11:32 am

Re: Bai Perron Test

Postby TomDoan » Fri Aug 12, 2011 4:17 am

iloverats wrote:dear
may you tell me how to choose the minspan value in baiperron.src?
thanks


That depends upon what you're trying to do. If you're using Bai-Perron as a specification test for a fixed parameter model, the smallest value for which they made a table is 10% of the sample size and it doesn't sound like they would recommend going smaller than that. If you're taking a breaking model as given and just want to find the best fitting values, the default value of the number of regressors is fine.
TomDoan
 
Posts: 2720
Joined: Wed Nov 01, 2006 5:36 pm


Return to Structural Breaks and Switching Models

Who is online

Users browsing this forum: No registered users and 1 guest