Page 1 of 1

breakpoint unit root test

Posted: Wed Apr 20, 2016 11:47 pm
by john1990
HI
I want to test for unit root with possible breakpoint .
the breakpoint is known. First I tested for unit root for two subsamples (before break date and after) Both results suggest that the null of unit root cannot be rejected , then I tested for the whole sample using breakpoint unit root test, but different specifications of the test gives very different result , for example if I choose INNOVATION OUTLIER the test rejects the null but for ADDITIVE OUTLIER option the null cannot be rejected , and i also get different results using Akaike or Schwartz criterion for lag specification. SO How can i decide what specification to chose ?
Here is the graph of my data , you can see the break at 2008

Re: breakpoint unit root test

Posted: Thu Apr 21, 2016 7:01 am
by trubador
Looks like you have a trending data with an innovation outlier that led to a break in both trend and the intercept. So it is quite normal for additive outlier specification to fail the test...