Error with panel data

[Willian Dash]

I currently deal with the data using fgls. I have a question that is there any problems if "estimated covariances" is equal to"number of groups". My instructor told me that "there is the repeated variables in my data, for example the variable gdp of export and import countries". Please advise. Thank you very much
Here is the descriptive:
. sum exportvalue gdp1 gdp2 lpi1 lpi2 elpi1 elpi2 regulatoryquality politicalstability distance commonborder

Variable | Obs Mean Std. dev. Min Max
-------------+---------------------------------------------------------
exportvalue | 540 3.69e+09 7.72e+09 191.739 5.28e+10
gdp1 | 540 2.56e+11 2.58e+11 9.91e+09 1.12e+12
gdp2 | 540 2.56e+11 2.58e+11 9.91e+09 1.12e+12
lpi1 | 540 3.021099 .5229104 2.06725 4.3
lpi2 | 540 3.021099 .5229104 2.06725 4.3
-------------+---------------------------------------------------------
elpi1lpilci | 540 39429.11 53627.14 8418.875 276955.3
elpi2 | 540 39429.11 53627.14 8418.875 276955.3
regulatory~y | 540 .0354467 .9635217 -2.244581 2.22636
politicals~y | 540 -.0812727 .8993196 -2.211743 1.477469
distance | 540 1637.217 719.7744 314.706 3089.79
-------------+---------------------------------------------------------
commonborder | 540 .2444444 .4301558 0 1

And here is the result of FGLS:

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: heteroskedastic
Correlation: common AR(1) coefficient for all panels (0.7488)

Estimated covariances = 90 Number of obs = 540
Estimated autocorrelations = 1 Number of groups = 90
Estimated coefficients = 9 Time periods = 6
Wald chi2(8) = 2056.81
Prob > chi2 = 0.0000

------------------------------------------------------------------------------------
lnexp | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------------+----------------------------------------------------------------
lngdpex | 1.11632 .0443255 25.18 0.000 1.029443 1.203196
lngdpim | .8699424 .0348179 24.99 0.000 .8017005 .9381842
lndis | -1.235037 .0858691 -14.38 0.000 -1.403338 -1.066737
lnelpi1 | .4750061 .0458459 10.36 0.000 .3851498 .5648625
lnelpi2 | .1907061 .0524039 3.64 0.000 .0879963 .2934159
regulatoryquality | .1550171 .054814 2.83 0.005 .0475837 .2624505
politicalstability | -.1463815 .0433558 -3.38 0.001 -.2313573 -.0614058
commonborder | .6490318 .1101282 5.89 0.000 .4331844 .8648792
_cons | -28.31576 1.436243 -19.72 0.000 -31.13074 -25.50077
------------------------------------------------------------------------------------

[Carlo Lazzaro]

Willian:
welcome to this forum.
Stata is telling you that you have N=90 and T=6.
That's it.
I cannot see where the mistake/error lurks here.
That said, as you're dealing with a N>T panel dataset, you should consider -xtreg- (provided that your regressand is continuous).

[Willian Dash]

Thank you for your information, Carlo.
I just wonder if it is a problem or not when my database has repeated values.
Because I examined trade intra-ASEAN, so my database will have repeated value, for example:

countryex	countryim	Time	lnexp	Export value	lngdpex	GDPex	lngdpim	GDPim
Brunei Darussalam	Cambodia	2010	11.01614	60848.5	23.28781	1.30E+10	23.2702	1.28E+10

Cambodia

Brunei Darussalam

2010

10.18271

26442

23.2702

1.28E+10

23.28781

1.30E+10

I conducted FEM, REM and Pool OLS. After testing, FEM is the most suitable model. And my data has autocorrelation and heteroskedasticity issues. How can I deal with it?
Please help advise. Thanks for your help.

[Carlo Lazzaro]

Willian:
if Stata complains about that, you can simply -xtset- your dataset with -panelid- only.
This trick comes at the cost of making time-series operators, such as lags and leads, unvailable, but still allows you to add -i.timevar- in the right-hand side of your regression equation.
Use cluster-robust standard errors to deal with both heteroskedasticity and epsilon autocorrelation.

[Willian Dash]

Thank you for your support, Carlo