panel regression with country and period fixed effects