Panel Unit Root testing questions! Help!

[leolii]

Hi, I'm doing my thesis report, I have a panel data, to have the panel Unit root test, I choose Hadri,Im, Pesaran and Shin and Levin, Lin & Chu methods. But while dealing the result, I have trouble to get the conclusion, all the variables have less than 5% P-value. Besides,Im, Pesaran, Shin W-stat and Levin, Lin's Null Hypothesis are both Unit root. But Hadri's Null Hypothesis is Stationarity. What can I get if I can't reject any of them. More than that, I can't even get the critical value when testing variables, anybody knows why?
Thanks

[ardinea]

Hi,

My thesis before also used panel data. for unit root testing in panel, you should choose the test that you prefer out of the several tests available. In my thesis before, in selecting the test result, I rely on the result of ADF Fisher Chi square test. I prefer ADF Fisher test because it does not require a balanced panel unlike the test for (IMPS). Also, the Fisher test can use different lag lengths in the individual ADF regressions and can be applied to any other unit root tests (Baltagi). you can read more in econometric text book by Baltagi "Econometrics Analysis of Panel Data". You can see it on the attachment below. If you use Eviews, you can read in Eviews help to find some information about panel.

If you are familiar with eviews, eviews will have a shorter output that gives an easier interpretation. good luck!!

[Aijan]

Hi there,

1. I have an unbalanced panel model and I need to check it for stationarity. So, I need to perform a Unit Root test. But I am a bit confused whether (1) I need to perform a panel unit root test separately for each variable I have in my model; or (2) I need to perform a kind of a combined panel unit root test for all variables in the model (maybe including them into a group or something like that...).

In case if I need to perform (2), how should I make it in eviews? When I open several variables as a group, I do not see a "Unit root" option under "Views" tab.

And then if I have a unit root, how can I make the model stationary?

2. And one more question: Should I perform a separate unit root test for a dependent variable or should I perform a common unit root test including both dependent and independent variables (again, group them or do something like that)?

P.S. I have a structured panel workfile and I am using EViews8.

Thanks a lot!