Lee and Strazicich

For questions that don't fall into one of the categories above, such as working with the RATS interface, using Wizards, etc.
Zankawa
Posts: 65
Joined: Mon Jun 15, 2015 2:23 pm

Lee and Strazicich

Unread post by Zankawa »

Hi,
I tested a series for unit root with structural breaks using the Lee and Strazicich test, and the output of the test is shown below

Lee-Strazicich Unit Root Test, Series LRAGR
Regression Run From 1971:01 to 2014:01
Observations 44
Trend Break Model with 2 breaks
With 1 lags chosen from 5

Variable Coefficient T-Stat
S{1} -1.0741 -6.4666
Constant 0.1453 5.3903
D(1982:01) 0.0452 0.7262
DT(1982:01) -0.1879 -5.0383
D(2002:01) 0.0664 1.0619
DT(2002:01) 0.0525 2.1996

My question is; how do I test for the significance of the break dates(i.e. 1982 and 2002) using the T-statistics provided. I am asking because when structural breaks are present, I may need to include dummies in the model to correct the structural breaks. I am using a multivariate model with several variables. As a result, it important that I know which breaks are significant so I can include the relevant dummies in my regression.
Thank you
TomDoan
Posts: 7814
Joined: Wed Nov 01, 2006 4:36 pm

Re: Lee and Strazicich

Unread post by TomDoan »

If you read through the thread on the procedure, there are three different questions which are basically the same as yours. This is a test for unit roots, not a test for breaks. The break points are picked to be least favorable to the unit root hypothesis, not to be the "best" breaks. The procedure doesn't even aim to produce a usable "alternative" model since it's an LM test and thus is calculated under the null.
Post Reply