Model specification and interpretation: A reply to Greenberg