Page 1 of 3
Quandt-Andrews Test Error
Posted: Thu Feb 14, 2013 12:53 pm
by kevinliberty
I am pretty new to EViews so I apologize if this is a simple fix.
I am building seasonally dependent forecasting that takes samples from November 2nd to March 1st of every year, so for example my sample ranges will look like this 11/02/2002 4/31/03 11/02/03 4/31/04...
I use the HAC (Newey-West) covariance option when building my regression estimation, which runs fine. I then go to run my Quandt-Andrews Breakpoint Test and I receive a message saying I have an illegal sample date, it is always an error between the 2nd and 3rd variable in sample range. Why doesn't the QA test allowed for interval sample ranges in this test?
Thanks
Re: Quandt-Andrews Test Error
Posted: Thu Feb 14, 2013 1:03 pm
by EViews Gareth
What's the build date of your copy of EViews? (Available from Help->About EViews).
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 6:55 am
by kevinliberty
Standard Edition - Jul 6 2012 build
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 7:32 am
by EViews Gareth
You should update to the latest version, and if that doesn't fix it, provide your workfile.
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 9:15 am
by kevinliberty
I have the Jan 02 2013 build now is that the most recent build?
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 9:37 am
by EViews Gareth
yes
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 10:08 am
by kevinliberty
Attached is the data, included in the spreadsheet is my formula and data sample range.
Thanks,
Kevin
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 10:39 am
by EViews Gareth
Worked for me.
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 10:56 am
by kevinliberty
Great thanks, I used the command lines to perform the steps and it worked fine.
Thanks for your help!
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 11:14 am
by kevinliberty
Wait I still am getting the same problem,
When I go to include the HAC Newey-West covariance option. I get the the sample error prompt.
Re: Quandt-Andrews Test Error
Posted: Fri Feb 15, 2013 12:32 pm
by EViews Gareth
Ah yes, my bad, I forgot to do HAC. I get the same message you do. It is a bug. We will fix it.
Re: Quandt-Andrews Test Error
Posted: Tue Feb 19, 2013 12:42 pm
by kevinliberty
Is there a tentative date when this bug we will be fixed?
Re: Quandt-Andrews Test Error
Posted: Tue Feb 19, 2013 3:59 pm
by EViews Gareth
Today's patch should have fixed it.
Re: Quandt-Andrews Test Error
Posted: Wed Feb 20, 2013 8:13 am
by kevinliberty
Excellent thanks the patch worked!!
I am having a problem with matching the results from EViews 6 to EViews 7.
The earlier iteration of the Quandt-Andrews reports these results (run on EViews 6):
Statistic Value Prob.
Maximum LR F-statistic (2/12/2011) 13.44626 0.1333
Exp LR F-statistic 4.052823 0.131
Ave LR F-statistic 4.979023 0.2444
Note: probabilities calculated using Hansen's (1997) method
The same test run on Eviews 7 reports this result:
Statistic Value Prob.
Maximum LR F-statistic (2/12/2011) 13.44626 0.0
Exp LR F-statistic 4.051269 0.0
Ave LR F-statistic 4.978227 0.0
It looks like there is a problem with the probability results for this test, as they should be identical.
Thanks.
Re: Quandt-Andrews Test Error
Posted: Wed Feb 20, 2013 8:15 am
by EViews Gareth
Not with HAC errors.