Am 01.03.2015 um 17:37 schrieb Mark Scerri:
HI,
I am rather new to time series analysis. I am trying to perform a unit
root test. I have been advised to perform the test on three cases.
Case 1: no constant term or time trend; True process is a random walk.
Case 2: constant term but no time trend: true process is a random walk.
here the constant term is only needed for the alternative hypothesis
Case 3: constant term but no time trend; true process is a random
walk
with drift.
here the constant is also needed for the null, because with a unit root
it cumulates to a drift (=linear time trend term).
But this case makes no sense really for testing. Because if you specify
the null as containing a drift/trend, it is almost surely nonsense to
not allow a trend under the alternative (trend stationarity). So are you
sure that's really what you were asked to do?
-sven