BaiPerron (Bai-Perron breakpoint test - Requires R)

[EViews Gareth]

This thread is about the BaiPerron Add-in that performs the Bai-Perron (1998) breakpoints test, as implemented in the R package "struccchange". Note R is required for this add-in.

Change Log

• 2010/03/24 - Initial release
• 2010/04/09 - Fix for bug caused by the global option setting for sending variables over to R
• 2010/04/12 - Added Log-likelihood, RSS and number of coefficients to the output for each set of breakpoints

[EViews Gareth]

New update posted, fixing an issue where users who had changed their global options to send over series to R in Upper Case would have problems.

[EViews Gareth]

Added Log-likelihood, RSS and number of coefficients to the output for each set of breakpoints

[kpukthua]

Hi,

Does anyone know how I can perform this test? I'm using Eviews 7 and can't find it.

Thank you.

[EViews Gareth]

You need to have EViews 7.1, and then simply download the add-in (linked to in original post)

[Merih]

I am running EViews 7.1 and have installed the BaiPerron add-in, but after the first pop-up for minimum sub-sample size, it displays the error message "R is not installed, please check your EViews documentation for instructions on installing R." I have R-2.11.1 installed in the Program Files directory. What should I do? Also, where can I find EViews documentation for 7.1?
Thank you very much.

[EViews Gareth]

You should read the manuals available from the help menu in PDF format

[kpukthua]

Merih: I also have the same problem. Can you please let me know once you figure out? Thanks!

[kpukthua]

Hi Gareth or anyone,

I successfully added in the Bai-Perron test. Do you know where I can get the manual to understand the results? I searched from Internet and help function in R program but couldn't find it.

I regressed adjusted R-square on time and got the following results. Will appreciate very much if anyone can help me explain the results. Thanks!


Bai-Perron breakpoint test
Date: 07/04/10 Time: 14:07
Sample: 1 3589 IF ARGENTINA=1
Included observations: 172

Breakpoints 0 1 2 3 4 5
BIC 190.7279 191.3778 197.3058 209.0713 222.1982 239.7879
Log-Lik -87.64269 -80.24641 -75.48920 -73.65068 -72.49291 -73.56649
RSS 27.90256 25.60315 24.22533 23.71293 23.39584 23.68973
N. Coefs 3.000000 6.000000 9.000000 12.00000 15.00000 18.00000

Chosen number of breaks: 0

[EViews Gareth]

I think it is fairly self-explanatory. The test found no breaks.

[Wira]

Hi Gareth and everyone,

I found the BP test can work with AR(1) equation, is it right? I found a paper Atkins(2002) that using AR(1) model. Is it different algorithm?
Can we use equation with trend such as y=a0+a1@trend ? OR only a constant as a proxy of the mean such as y=a0 ?
Please enlighten me. Thanks.

[EViews Gareth]

ARMA terms are not supported.

[Wira]

Dear Gareth,
What about the equation with constant and a trend? Is it in line with Bai Perron (1998) algorithm or their theory?
Thanks.

[EViews Gareth]

No idea. The add-in just calls R code and doesn't do any computations itself. You should check with the R documentation.

[EViews Glenn]

Simple trends violate the regularity conditions. Polynomial trends of a particular form are allowed.

See Bai, J., 1997a. Estimation of a change point in multiple regression models. Review of Economic and Statistics 79, 551-563. There is also a discussion in Perron's survey article "Dealing with Structural Breaks", Palgrave Handbook of Econometrics, Vol. 1: Econometric Theory, K. Patterson and T.C. Mills (eds.), Palgrave Macmillan, 2006, 278-35.