Page 1 of 1

APGRADIENTTEST - general Andrews-Ploberger tests

PostPosted: Tue Nov 30, 2010 12:16 pm
by TomDoan
This is a procedure for applying the Andrews-Ploberger change point analysis to a more general type of model than the linear regression than the @APBreakTest procedure allows.

apgradienttest.SRC
(9.21 KiB) Downloaded 169 times


The input to this are the series of gradients from maximum likelihood estimation. These can be fetched using the DERIVES option on MAXIMIZE or GARCH. For instance, the following would do a break test on the GARCH(1,1) model estimated in the GARCHUV example:

Code: Select all
garch(p=1,q=1,hseries=hh11,derives=dd) / dlogdm
@apgradienttest(graph)
# dd

Re: APGRADIENTTEST - general Andrews-Ploberger tests

PostPosted: Tue Nov 30, 2010 7:00 pm
by soju
Hi Tom
Thanks so much for the procedure! I just had a bug running it though:
I run an ACD model using maximize and fetch the derivatives dd:
Code: Select all
maximize(method=bhhh,recursive,iterations=150,derives=dd) gln 3 *
@apgradienttest(graph)

as soon as I hit enter after the last line, it appears Rats32s.exe program has stopped working and I have to force to quit.
Also when I run the porcedure WITHOUT the "graph" option, it's not crashing but appears the following:
## OP3. This Instruction Does Not Have An Option TIT
>>>>ction=define,title=<<<<




Regards

Re: APGRADIENTTEST - general Andrews-Ploberger tests

PostPosted: Wed Dec 01, 2010 10:43 am
by TomDoan
Replace:

report(use=areport,action=define,title="Andrews-Ploberger Break Tests")

with

report(use=areport,action=define)

The TITLE option on REPORT was added with version 7.3.

Re: APGRADIENTTEST - general Andrews-Ploberger tests

PostPosted: Wed Dec 01, 2010 6:08 pm
by soju
great it works now. Thank you so much