st: dfuller test statistic

[Skipper Seabold <jsseabold@gmail.com>]

All,

In Stata 11, when I do

webuse lutkepohl2
dfuller ln_inc, lags(4) regress

The reported test-statistic for a unit root is just the t-statistic on
the lagged ln_inc.  However, the Hamilton textbook and MacKinnon
(1994) recommend that when lagged values of the difference are
included that the appropriate test statistic is

Z = (T/(1-sum(coefficients on lagged differences))) * (coefficient on
lagged level - 1)

where T is the number of observations (not including lags).

Is this just a recommendation?  Is the single t-statistic valid to
test, and if so are the regression surface results of MacKinnon (1994)
with no lagged differences valid for this specification?

Thanks,

Skipper
*
*   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/