Sampling for weight changes in animal populations with or without identification

Abstract

Procedures for studying weight changes in animal populations by sampling on successive occasions, may or may not involve identification of the animals sampled. Since changes in a population are in general more precisely estimable from animals common to both samples, identification of the common animals enables a change in the mean weight of the population to be estimated more precisely, since then the submeans of the common animals become available for utilization in the construction of the estimators. The extent to which animals are identified during the two weighings determines whether two, one, or none of these submeans are ascertainable from the data. Accordingly there are three different procedures for estimating a change in the population's mean weight with minimized variance, one appropriate for each case. To enable comparison, the variances of these three estimators are tabulated as a function of the number of animals common to both samples and the correlation between successive weight measurements. The effect of various weighing practices in farming and research on the precision of the best obtainable estimators of change are discussed. The general formulas developed for the variance in each case pinpoint the components of variation and enable calculation of the precision to be gained by manipulating these.