Investigating the impact of representation upon coarse-grained models