The effect of measurement bias on the design of samples for the estimation of morbility statistics