[Gretl-users] On the use of Fixed Effect and Random Effects models for Panel Data

I have read a blog stating that if there is heterogeneity in the panel data the appropriate model is either the Fixed Effect or Random Effects model, not the OLS Regression. Do you agree with the blogger? I'm interested to know to determine the appropriate panel regression model when there are autocorrelations, heterogeneity, multicollinearity, and non-normality. Appreciate if you suggest a textbook that deals with the OLS, FE, and RE models for panel data. Thank you. Lema