Unit root test

[Effiezal]

Hi everyone

I need some help in understanding unit root test. Just find out from
Eviews that I might have to run unit root test to check whether the
series is stationary or not.

I have run it on eviews using panel unit root test and ALL test showed
that I can reject the null, which mean that the series is stationary,
correct?

This is when I got confused.

Anyway, another question :

Is unit root a CUMPOLSORY test before running a rgression? I have done
all the other stuff, multicollinearity, autocorrelation, normality
etc...but what about unit root?

Cheers

Efi.

[Raquel Guimarães]

Hi Effiezal,

 

Yes, if you rejects the null hypiotesis, the series is stationary (see below an eviews's help fragment):

 

Regression will calculate the t-statistic for this hypothesis, which may be used to test the significance of  g. The t-statistic, however, cannot be referred to the critical values in the standard t table, as reported in the regression output, since under the null hypothesis the left-hand variable is nonstationary and this table no longer applies.
 

About the question of the compulsority of a unit root test, I think that it's crucial because if the dependent variable and the explanatory variable are moving together, they, in fact, are not explaining eachother, and a high R-squared should be refleting this in order to measure the explain's power of the model (this problem is called spurious regression).

Hope this little can help u...

Regards,

Raquel
 

2006/3/16,  <Effieza...@gmail.com>:

rgu...@face.ufmg.br

[Effiezal Wahab]

Thanx Raquel

With regards to unit root test:

I have done unit root test for all of my variables, including the
endogenous variables,and find non of them are non-stationary. So, it
is okay right, for me to run regressions?

What bout 1st difference and 2nd difference? What is that ?

Sorry to bother you.

Efi.

--
Effiezal Aswadi Abdul Wahab
PhD Student
Department Accounting and Finance
The University of Auckland Business School
Private Bag 92019
Auckland
New Zealand

0193866157 (Mas)
0211294028(New Zealand)
09-3737599 x 86507(office)
Visit us at : http://www.business.auckland.ac.nz/

http://effiezal.blogs.friendster.com/my_blog/

[Raquel Guimarães]

yes... this is ok! now you can run four million regressions.... :P

1st difference means when you have a non-stationary series and if you apply to it a transformation,  i mean, y(t)-y(t-1), the new series (i'll call it w(t)) can be stationary.

if the new series again it's non-stationary, you shall apply a second difference, that is, w(t)-w(t-1) or [y(t)-y(t-1)] - [y(t-1)-y(t-2)].

Some good reference books on this:

PINDYCK, Robert S; RUBINFELD, Daniel L. Econometric models and economic forecasts. 3. ed. New York: McGraw-Hill, c1991. 596 p ISBN 0071008667 (broch.)

BOX, George E. P; JENKINS, Gwilyn M. Time series analysis : forecasting and control. rev. ed. Oakland: c1976. 575p. 1v. ISBN 0816211043 (enc.)

2006/3/16, Effiezal Wahab <effieza...@gmail.com>: