TAR Panel Unit Root Tests - with bootstrap

[j.galimberti]

I have developed some program codes for TAR panel unit root tests, extending Caner and Hansen (2001) to panel data context, with a Fisher-type test proposed by Maddala and Wu (1999). This code is available at:

http://sites.google.com/site/jkgeconoen ... ical-works

The code collection also contains a procedure to bootstrap Panel Unit Root p-values (no threshold).

The working paper describing the procedures may be found at:

http://ideas.repec.org/p/pra/mprapa/14237.html

I will be glad to receive any comments. In the paper, I haven't tested all the features that I intended to allow the code to do. So, I am also open for bugs reports.

[equn]

hi galimberti, thank you for your codes!
btw, do you have any idea about how to code LLC test and IPS test in eviews? they are much more popular in panel unit roots test.

[j.galimberti]

Hello Mr. Wang,

I am glad to hear from you.
Well, for the linear case (no threshold) the LLC, the IPS, and even the MW are ready to use directly by eviews interface. You will only need to code if you want to bootstrap your p-values, and that's what I have done for MW...I haven't done for LLC and IPS, but it wouldn't be that hard...I'm a little out of time now, but perhaps in the future we could we think better about it...

For the threshold case the thing is a little more complicated...the MW (or Fisher-type) facilitates a lot cause there's no need to assume a common autorregressive coeficient, and each process can be estimated (almost) independently. I think of extending the procedures for IPS and LLC in the future, cause I haven't seen this in the literatura until now...but there some theorethical efforts are needed.

Sorry if I didn't help you so much...I have to study for a macro exam next Monday!!!
If you have MSN Messenger add me and we can talk fastly: jakaga2002@yahoo.com.br