Page 1 of 1

Quandt-Andrews breakpoint test for a generated variable

Posted: Sun Jan 12, 2014 1:03 pm
by wetkneehouston
Hi,
I have generated a series of 1500 observations like this:

Y1 = 0+sqr(3)*nrnd for sample 1 - 500
Y1 = 50+sqr(3)*nrnd for sample 501-1000
Y1 = 50+sqr(30)*nrnd for sample 1001 - 1500

As part of a question, I am asked to perform the Quandt-Andrews breakpoint test and find the observation where the break occurs. As I understand it, you need to estimate an equation to use this test, but all I have is the variable Y1. To perform this test do I have to make this variable into an equation somehow? How would I go about doing this and performing the test? Any help would be really appreciated!

Re: Quandt-Andrews breakpoint test for a generated variable

Posted: Sun Jan 12, 2014 2:41 pm
by EViews Glenn
Presumably you'd want to regress on a constant

Re: Quandt-Andrews breakpoint test for a generated variable

Posted: Sun Jan 12, 2014 2:43 pm
by trubador
As a programme code, you can do it as follows:

Code: Select all

wfcreate u 1500 series y smpl 1 500 Y = 0+sqr(3)*nrnd smpl 501 1000 Y = 50+sqr(3)*nrnd smpl 1001 1500 Y = 50+sqr(30)*nrnd smpl @all equation eq.ls y c eq.ubreak 15
Please lookup the EViews' help system for "Andrews-Quandt breakpoint test" for more details.

Re: Quandt-Andrews breakpoint test for a generated variable

Posted: Sun Jan 12, 2014 3:27 pm
by wetkneehouston
As a programme code, you can do it as follows:

Code: Select all

wfcreate u 1500 series y smpl 1 500 Y = 0+sqr(3)*nrnd smpl 501 1000 Y = 50+sqr(3)*nrnd smpl 1001 1500 Y = 50+sqr(30)*nrnd smpl @all equation eq.ls y c eq.ubreak 15
Please lookup the EViews' help system for "Andrews-Quandt breakpoint test" for more details.
Thanks a lot for your help, that seems to work!