Relative efficiency of ratio estimates of animal disease prevalence

Abstract

An evaluation was made of the efficiency of the ratio estimator of the prevalence of animal disease relative to the unbiased estimator to determine when ratio estimation could be expected to offer an advantage. Actual data from a mastitis survey, in which all cows-in-milk in randomly selected herds were tested for the presence of various pathogens, were used to make the evaluation. Correlations of the numerator and denominator variables of the ratio estimate were found to be consistently too low to expect these estimates to be competitive with the unbiased estimates. Yet, the ratio estimates were found to be as efficient as those of the unbiased estimates because the coefficients of variation of the numerator and denominator variables differed sufficiently. The conditions under which ratio estimation will perform well are specified.