My name is Daniel Ventosa-SantaulÓria. During the last years I have noticed that many practitioners perform unit-root tests, such as the ADF, without really studying the deterministic components; they include constants and trends without being fully aware that, in the presence of a unit root, the latter become a linear trend, and a quadratic trend, respectively. It is important to recall that deterministic trends tend to asymptotically dominate stochastic ones. Along with a colleague (M. Gˇmez), I developed a test specifically designed to test for the presence of a drift (and a possible shift in the drift) once there is evidence of unit root (recently published at the Journal of Time Series Econometrics, :o) ). Such test could be considered as a companion of any unit root test.
My proposal is the following: I would
like/love to include the test in GRETL as a Unit-Root companion test. I don't know if my proposal is too pretentious or plainly unrealistic. Although the test has been revised by experts and published in a fine journal, maybe the test is not important enough.
The fact is that I programmed the test using matlab code (the test is actually extremely simple; it is a matter of simple LS regressions) but I am unable to write it in C language. It is my first trial in a gnu/open source project and I would feel very proud if I could collaborate in such a great program as GRETL. If there is a chance to do all this, please follow this link
to read the working paper version of the test. The published version can be found here
and a preliminary
matlab code of the test can be found here
. (I prepared a much better one but I have not yet uploaded it). Many thanks in advance for your attention; I look forward to hear from you,