On the estimation of the proportion of positives in a sequence of screening experiments