Hi guys,
I have a panel data set with 2Ts 500N and 2Ts 30N
I use OLS and Fixed effects estimations
I cannot test for autocorrelation but I can heteroskedasticity
Should I use the heteroskedastic tests to determine whether I need clustered standard errors, or should I just cluster anyway because of having 2 Ts.
I would use clustered errors for OLS estimations and robust errrors for FE - should i only use them if heteroskadicity is present or use them even if data shows homoskedasticity?
Also if I should just uses the errors regardless, does this apply to both OLS and FE?
Thank you