Page 5 of 6

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Wed Sep 05, 2012 4:48 am
by AndronikiK
zivot_andrews.wf1
(9.77 KiB) Downloaded 821 times
To whom it may concerns,

I would like to introduce a problem that i confront with the implemtation of the ZA test on my time series. First of all, i have the E-views 7 software. The problem appears when i try to apply the ZA test for my annual data concerning the world's GDP (constant 2000 US$). Despite the fact that i get results for both intecept and trend, i dont get any results for intercept and trend separately whatever the lag length. The message that appears on my screen every time is "Near singular matrix error. Regressors may be perfectly collinear".

I would appreciate if any help could be available since i need the results for my dissertation.
You can find attached the workfile with the GDP in levels and in logarithms.

Thank you in advance,
Niki

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Wed Sep 05, 2012 5:55 pm
by mrrox
Gareth,

I get different results for my data. When I run this test in Eviews, I get a break which is different from the break obtained when this test is run in R on this data.

I used general to specific method to determine the lag length. But Eviews uses AIC to set the lag length.

And in most cases, I get near singular matrix error message.

Anyway, hope you will look into it.
mrrox

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Wed Sep 05, 2012 6:58 pm
by EViews Gareth
You'd have to ask Trubador.

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Thu Sep 06, 2012 9:08 am
by bvguizar
Hi Gareth,

Sorry for the delay, please find file attached

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Fri Sep 07, 2012 10:55 am
by bvguizar
It said 1 to 177 but can you help me now? I have uploaded the file correctly.


Hello

Thank you for your reply


I have two questions about the Zivot-andrews unit root test. How do you choose the perfect lag for it? I am using 12 because I have monthly information. I am working with 177 monthly observations and 5 variables using trend and constant. I know it has structural breaks and I am trying to identify them. However, I am not sure if I am identifying the breaks correctly since I used the maximum lag of 12 and I got an error on le85nst.

The ones I want to check with Z-A unit root test are:

le85nst
le85p
le85q
lgp
ly

The series I get the error is the one named le85nst. The error I get from this serie is Near singular matrix error: Regressors maybe perfectly collinear.

The original series is e85nst, I added logarithms and converted in le85nst and after that I used the seasonal adjustment of X12 since these are american series.

I will use a dummy to reflect the breaks in my ARDL/VAR model.

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sat Sep 08, 2012 9:35 am
by trubador
bvguizar wrote:The series I get the error is the one named le85nst. The error I get from this serie is Near singular matrix error: Regressors maybe perfectly collinear.

Use the code here: viewtopic.php?f=15&t=1432#p4925

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sat Sep 08, 2012 1:08 pm
by bvguizar
Hello Trubador,

I get the error attached

do you know why I might get it?

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sat Sep 08, 2012 3:51 pm
by trubador
Just change the initial part of the code:

Code: Select all

call zivot(le85nst,"C",12)

And then you should be getting the following output:

Variable(s) LE85NST
t-stat(s) -3.618547
Lag(s) 1.000000
Break 2000M01
DU1 p-value 0.022781

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sat Sep 08, 2012 5:38 pm
by bvguizar
Hi,

with

call zivot(le85nst,"C",12)

I got

Variable(s) LE85NST

t-stat(s) -5.055183
Lag(s) 10.00000
Break 2005M12
DU1 p-value 0.000815

Is this wrong?

The file I originally updated here got corrupted then I re did it as this one but the values have changed I cannot get why is the same info as the one I uploaded before.

Could you please help?

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sun Sep 09, 2012 8:15 am
by trubador
Nothing seems wrong with your second workfile and when I re-run the code I get the same results as before.

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sun Sep 09, 2012 3:15 pm
by bvguizar
Hi,

Thank you so much for your help, I doubled checked.

So this break found is just for constand and trend? Or how does it work?

Variable(s) LE85NST

t-stat(s) -3.618547
Lag(s) 1.000000
Break 2000M01
DU1 p-value 0.022781

How do you interpret the above results?

I am working with 177 monthly observations and 5 variables and I want to check the model with constant, trend and constant and trend. I know the series have structural breaks and I am trying to identify them and use a dummy where is needed.

Could you please help me?
Thanks
Blanca

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Sun Sep 09, 2012 11:27 pm
by trubador
Please refer to the original study in order to better understand the approach and the details on different specifications. Other than that, I'd like to remind you that analyzing structural breaks within a modeling context is a different task. Although ZA unit root test will provide you some useful insight, you need to use a more general approach (e.g. Bai-Perron) if you are planning to build a model using these variables.

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Mon Sep 10, 2012 2:19 pm
by bvguizar
Dear Trubador

Do you mean this study:

'Zivot-Andrews Unit Root Test
'Reference: Zivot, E. and Andrews, D. W. K. (1992), “Further Evidence on the Great Crash, the Oil-Price Shock, and the Unit-Root Hypothesis”, Journal of Business & Economic Statistics, Vol. 10, No. 3, pp. 251-270.

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Tue Sep 11, 2012 4:08 am
by standakostka
Hi,

thanks for the add-in, it looks really easy to use. Can be this test performed directly on for example gdp growth rates series or should there be regression first done?

Re: ZAURoot (Zivot-Andrews Unit Root test)

Posted: Fri Dec 21, 2012 7:33 am
by bvguizar
I am trying to use this add in on my lap top but it just doesn't work. I have eviews 7 and it doesn't let me use it on it. It marks an error.

Someone help please