First of all, Pesaran's (2004) test for cross-sectional dependence is the right test to use in the presence of unbalanced panel data, as the test statistic has been modified to account for the fact that residuals for subsets of each time period are not necessarily mean 0. Therefore you should implement "xtcsd, pesaran".

Second of all, if cross-sectional dependence is found, you should ideally use Driscoll-Kraay standard errors, which handle arbitrary forms of spatial dependence in conjunction with heteroskedasticity and autocorrelation. These are implemented with the stata command xtscc (ssc install xtscc). Note that these standard errors have worse coverage rates than clustered standard errors (by N), if cross-sectional dependence is not present. 

Hope this helps!

I recommend to answer first question to use  Pesaran's (2004), In Stata command  you should implement "xtcsd, pesaran".; and for second question I recommend to use Driscoll-Kraay standard errors,

xtcsd, pesaran

unknown function *sqrt()

i found error while using pesaran comand in stata

why this happens?

xtcsd is a user-written code. You have to install it first. Try the following command in stata:

findit xtcsd

Flagging Victos answer, this works only for regularly spaced data. What if the data are like 1991 2001 2011 etc?

Dear Farheen

You may go through the material using below link;

http://fmwww.bc.edu/RePEc/bocode/x/xtcsd.html

see the link below

https://www.statalist.org/forums/forum/general-stata-discussion/general/1352795-new-package-xtdcce2-and-xtcd2-available-on-ssc-to-estimate-common-correlated-effects-and-test-for-cross-sectional-dependence

@hugh I have applied three test for checking cross sectional dependence ( pesaran, frees and Friedman). Pesaran test indicates the presence of cross-section dependency, where as other two test indicates absence of cross sectional dependency. My N> T.

How to proceed further with these results?.