Model-based double clustering