heteroskedasticity

[Lovish shantanoo]

Hello.

im actually working on my dissertation which on the impact of dtt on FDI.

To start with, i test for correlation for my variable ( there were correlation among some of my iv)
- test for unit root ( variable are statationary)
-run my ols
-run the fixed effect (xtreg ,fe)
-run the random effect ( xtreg, re)
-hausmn test (hausman fe re sigmamore)- hausman select the fixed effect

should i test for cross section dependence , heteroskedasticity, autocorrelation?

[Carlo Lazzaro]

Lovish:
please (re) read the FAQ on how to post (more) effectively (by the way: what is dtt? and FDI?).
That said, if you have a large N, small T panel dataset (as -xtreg- would imply) and you suspect heteroskedasticity and/or autocorrelation, you should simpy robustify or cluster your standard errors.

[Lovish shantanoo]

dear carlo
my title is the impact of double taxation treaty on foreign direct investment
i have taken a sample of 12 SADC countries for the period 1990-2012.

[Carlo Lazzaro]

Lovish:
so you have a small N, large T panel dataset and -xtreg- is not the best approach; I would look at -xtpcse- and -xtgls- commands, instead.

[Lovish shantanoo]

im testing for cross sectional dependence for the fixed effect.
here are the results
Breusch-Pagan LM test of independence: chi2(66) = 115.530, Pr = 0.0002
Breusch-Pagan LM test of independence: chi2(66) = 115.530, Pr = 0.0002
Based on 22 complete observations

. xtcsd, pesaran abs


Pesaran's test of cross sectional independence = -0.354, Pr = 0.7231

Average absolute value of the off-diagonal elements = 0.219

i have no idea to interprete the result

[Carlo Lazzaro]

Lovish:
as per FAQ, please state where you got the user-writen programmes -xttest2- and -xtcsd- from. Thanks.
How to intepret the results of -xtcsd, pesaran- is covered in the -xtcsd- dedicated article that you can download free of charge from http://www.stata-journal.com/sjpdf.h...iclenum=st0113.
As per the seemingly conflicting results, they may depend on the fact that you seem to have a very limited number (22) of complete observations.