RE: st: Panel unit root test

[Nick Cox <n.j.cox@durham.ac.uk>]

The first difference can't be calculated for the first observation in each panel. So, the report of missings is not surprising. What else may be wrong I can't say. 

Nick 
n.j.cox@durham.ac.uk 


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Muhammad Anees
Sent: 21 November 2011 14:46
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Panel unit root test

The problem inheritedly is because you have missing observations in
your data. Before going to explore other issues, you must check of
creating such first differenced variables really is what you want.

On Mon, Nov 21, 2011 at 7:23 PM, Tong, Tingting <ttong2@utk.edu> wrote:
> Dear All,
>
> I am very new in Stata and I am going to use PMG method in my thesis.
>
> I have a question about panel unit root test. My data has 44 panels from 1958 to 2005.
>
> For the panel unitroot test, i use "help xtunitroot" in stata and it gives me some describtions. I used IPS and had the result like followings:
>
> . xtunitroot ips lnvadd, lag(aic 8) trend
>
> Im-Pesaran-Shin unit-root test for lnvadd
> -----------------------------------------
> Ho: All panels contain unit roots           Number of panels  =     44
> Ha: Some panels are stationary            Number of periods =     48
> AR parameter: Panel-specific                Asymptotics: T,N -> Infinity
> Panel means:  Included                                        sequentially
> Time trend:   Included
> ADF regressions: 1.14 lags average (chosen by AIC)
> ------------------------------------------------------------------------------
>                    Statistic      p-value
> ------------------------------------------------------------------------------
>  W-t-bar              4.6401        1.0000
> ------------------------------------------------------------------------------
> I think this means lnvadd has unitroot since p-value is very big.
>
> Then I want to see if i take the first difference of lnvadd, does it still have unit root or become stationary.
>
> I used generate code and it reported that 44 missing values. I am not sure if this the the right way to get the first difference data in Stata.
>
> generate lnvaddd1=d1.lnvadd
> (44 missing values generated)
>
> Then I run the unitroot test again for the d1.lnvadd variable i just generated.
> . xtunitroot ips  lnvaddd1, lag(aic 8) trend
> Im-Pesaran-Shin unit-root test for lnvaddd1
> -------------------------------------------
> Ho: All panels contain unit roots           Number of panels  =      0
> Ha: Some panels are stationary              Number of periods =     47
> AR parameter: Panel-specific                Asymptotics: T,N -> Infinity
> Panel means:  Included                                        sequentially
> Time trend:   Included
> ADF regressions: 0.00 lags average (chosen by AIC)
> ------------------------------------------------------------------------------
>                    Statistic      p-value
> ------------------------------------------------------------------------------
>  W-t-bar                   .             .
> ------------------------------------------------------------------------------
>
> It is obviously wrong since the number of panels become zero.
>
> I donot know where goes wrong. All I want to know is that the variable is not stationary in level, so I take first difference, see if they are stationary.
> If the data are stationary after taking first difference, then I want to say they are I(1), then I can go to next step to test the cointegration.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/