Nolan,
In the article written by the authors, they acknowledge that this test has power
problems. Maddala and Kim (1998) in their Unit Roots, Cointegration, and Structural Change
suggest that it should not be used at all. The size distortions with finite samples
of this test are serious according to MK(p.81). The PP test is less reliable than the ADF
test when a predominance of negative correlations inhere within first differences. It has
paltry power with negative MA.
The Elliot, Rothenberg, Stock test is better than both. However, Kit Baum has
maintained that the covariate adjusted ADF test may enhance the power from about
.12 to .58 (Baum, presentation at 2007? NAStata Users's Group meeting).
Either the ERS test or a CAADF might be better.
- Regards,
Bob Yaffeee
Robert A. Yaffee, Ph.D.
Research Professor
Silver School of Social Work
New York University
Biosketch: http://homepages.nyu.edu/~ray1/Biosketch2008.pdf
CV: http://homepages.nyu.edu/~ray1/vita.pdf
----- Original Message -----
From: Nolan Ritter <[email protected]>
Date: Wednesday, February 25, 2009 8:54 am
Subject: st: Phillips-Perron unit root test
To: [email protected]
> Greetings,
>
> I would like to employ the Phillips Perron test using the Stata
> command pperron on a time series of finance data. The output for this
> test, given on page 184 of Stata's time series manual, includes two
> test statistics: Z(rho) and Z(t). My main question is whether one test
> should be preferred over the other, and under what condition? A second
> question is whether the rho in the output corresponds to the alpha in
> the original article by Phillips and Perron (1988).
>
> Any insights would be greatly appreciated.
>
> Nolan
>
>
>
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/