Hi
There are two models.
Data taken from only 1 country and there are 25 institutes.
total observations=425
Years= 2003-2019

Hausman's selected the Fixed Effect Model in Model 1; and Random Effect Model in Model 2) (but facing heteroscedasticity issue)

There are three methods to remove the hetero issue in unbalanced panel:
  1. Robust Standard Error (but giving insignificant results)
  2. Transformation (Taking log) (i have taken the log but not giving more significant results 3/8)
  3. The last one is Weighted Least Square Method (GLS/FGLS) (giving good results) (so can we use this method for both fixed and random effect models??)
Many Thanks!