Cross section dependence test

[mehmed han]

Hello;
I try to do cross sectional dependence test in Stata. My purpose is making pretest in order to select correct type of unit root test. Since I have T>N, I want to use Berusch-Pagan (1980) LM test and Pesaran (2004) CD cross sectional dependence test.
However, I can't do that.
how can I do these tests?
I mean what should I do step by step?( Should I run any model before cross section test? or Is cross section dependence test applied on series?)
Also, what command should I use for these two tpe of cross section dependence tests?
Please help.

Thank you very much.
pc: I have dynamic panel model

[mehmed han]

Any suggestions for my question?

[mehmed han]

I think I wrote, I should answer.
As far as I know, Series cross section dependence test cannot be done Stata. It could be done Gauss 9 and its later versions program.
Cross section dependence test can be done both series and residuals. Before applying unit root test, it is usually done on series in order to determine what kind of unit test should be done.
For residuals,as I understand, it's done for diagnosis test.

[Jesse Wursten]

Hi, the cross sectional dependence test is implemented as xtcdf on ssc. You can install it by typing "ssc install xtcdf", you will find instructions on how to use it when you type "help xtcdf".

[mehmed han]

Hi, the cross sectional dependence test is implemented as xtcdf on ssc. You can install it by typing "ssc install xtcdf", you will find instructions on how to use it when you type "help xtcdf".

Hi Jesse;
xtcdf command is a postestimation command for cross section dependence test and one cannot use it in order to check cross section dependency of series.. I mean it is not a preestimation command and it is used for checking whether the residuals of a regression model have a cross section dependence or not. In short, it is for examining residuals, not for series cross section dependence.

[Jesse Wursten]

You can run xtcdf on any series you want, they don't need to be residuals (whether you would want to do that, I have no idea).

[mehmed han]

You can run xtcdf on any series you want, they don't need to be residuals (whether you would want to do that, I have no idea).

Jesse;
To use xtcdf command, don't you need to estimate regression model before to use xtcdf command?

[Jesse Wursten]

No, you can supply any list of variables to xtcdf, it doesn't require any estimation to take place. See "help xtcdf".

[Tasfaye Fikadu]

No, you can supply any list of variables to xtcdf, it doesn't require any estimation to take place. See "help xtcdf".

Jesse Wursten yes it does nor need estimation. However, this command is for cross-sectional dependence described in Pesaran (2004) and Pesaran (2015) which is appropriate for N>T. Berusch-Pagan (1980) LM test that is preferable when T>N was the question and at the same time I was searching for.