My study is the study of Bilateral trade relations: A Gravity Model analysis using Panel Data methodology.
log (Tradeijt) = 𝛼0+𝛼1log (GDPit*GDPjt) + 𝛼2log (PCGDPit*PCGDPjt) +𝛼3log (Distanceij) +𝛼4log (PCGDPDijt) +𝛼5 (TR/GDPit) +𝛼6 (TR/GDPjt) + 𝛼7(Borderij) + Uijt
My model for analysis.

1.I performed POLS, then perform diagnostic tests like for multicollinearity (no multicollinearity in POLS), Heteroskedasticity (it exists, so I regress POLS again for Robust Std. Errors), Normality test( regressors are not normally distributed, but residual is.)
2.Then I performed FE and RE regression, performed Hausman Test (Prob>chi2 = 0.0039 i.e did not accept H0).FE is appropriate.
3.Again FE regression with all the regressors which are being significant in FE.i.e xtreg logXijt logGDPitGDPjt logPCGDPitPCGDPjt TRGDPit TRGDPjt logDistanceij
4. Then heteroskedasticity test, as heteroskedasticity is present in FE so I did robust reg. Cross sectional Dependence is also there as well as Autocorrelation. I used command
xtregar logXijt logGDPitGDPjt logPCGDPitPCGDPjt TRGDPit TRGDPjt logDistanceij, fe for correcting auto correlation, but for cross sectional dependence I did nothing. So after Diagnostic checking, logGDPitGDPjt logPCGDPitPCGDPjt these two variables became insignificant too.
Please let me know the further analysis or corrections.