ZAURoot (Zivot-Andrews Unit Root test)

[mrrox]

Gareth and others,

Is it possible to run Ziwot-Andrews unit root test manually in Eviews?

I cannot export the add-in available for Eviews package as I have an older version. My uni has version 7 but I can gather that the add-in is only available for 7.1.

Can you please confirm if I can run this test manually in OLS?
Than you,
mrrox

[startz]

You might ask the university to update...it's free.

[mrrox]

You might ask the university to update...it's free.

I am going to ask the IT guys at the uni to update the package then.

On another note, can this test be run manually by creating dummy variables and entering the equation into OLS estimation menu and comparing the stats to the critical values tabulated by ZA. I know that I have to do this sequentially rather than defining the break myself in the model. If so I will do it manually as i seem to have more than one break in my data.

please reply.
mrrox

[trubador]

If you have multiple breaks in your data, then you should use other tests (e.g. Bai-Perron) that are specifically developed for the purpose. Such procedures are designed to locate and identify the breaks endogenously, so you can always use dummy variables and usual significance tests if you know the exact location and the properties of the breaks.

[mrrox]

after fruitless efforts to install za unit root test add in, I used R and ran the test. but that test did not locate the break dates correctly. some of my data have two or more breaks some have single break. I will run the Bai Perron test in R, just found the script for it.

[mrrox]

If you have multiple breaks in your data, then you should use other tests (e.g. Bai-Perron) that are specifically developed for the purpose. Such procedures are designed to locate and identify the breaks endogenously, so you can always use dummy variables and usual significance tests if you know the exact location and the properties of the breaks.

Hi turbador,

many thanks for the reply post!

Which significance tests do you mean?

Also, Bai-Perron test is not a unit root test, But I am using it anyway to locate the breaks and use dummies in the model. In many cases, it finds break dates correctly but in some cases it shows incorrect break dates-no break in the series but it indicates there is one. Do you know any multiple break unit root tests which either run on Eviews or R?

Thanks
mrrox