A debate on the classical and gravity trade models: Testing EU trade facts using direct inference