Why is it better to compare the per capita sick day rate among the low, medium and high populations than the total number of sick days?