When can Group Level Clustering be Ignored? Multilevel Models Versus Single-Level Models with Sparse Data