A selectivity model for fragmented relations: Evaluated for different standard data distributions