Dear all, I estimated a fixed effect selection model and got insignificant Rho coefficient. Does the lack of a statistically significant correlation among the errors invalidate the model? The random effect counterpart of the model has a significant rho coefficient. Does this imply that the RE is preferable than the FE? Thanks, Daniel