The precision for noticing significant differences among relativefrequencies increases as sample size increases. Although significantdifferences between observed prevalences can be estimated by theoverlap of confidence intervals, this ability to identify statisticallysignificant differences in trends could be improved by adding auto-mated statistical tests (e.g., chi-square test). In large herds it maynot be feasible to score every cow on a regular basis. These farms canrandomly select a proportion of the herd to be scored at regular timeintervals to obtain an accurate representation about the impact ofDD on the cows on a particular farm. The number of cows thatneed to be scored per farm results in a sample size that depends ona number of factors (e.g., lowest M-stage prevalence, the precisiondesired, total population size, the accuracy of the scorer, scoringtechnique) (Humphry et al., 2004; Thrusfield, 2005; Naing et al.,2006). It is important to minimize large changes in the populationscored over time. Large changes in the population scored duringdifferent scoring events can artificially cause significant changesin proportions that are not representative of the herd level situ-ation. In addition, not maintaining a consistent population to bescored decreases the amount of animals with consecutive scorescontributing to the transition matrix and therefore, the predictions.