Gill J L, Ericsson G F, Helland I S
Biometrics. 1986 Dec;42(4):981-7.
In counting internal helminthic parasites (the "worm burden") of domestic animals, physical restrictions often lead to sampling by small aliquots of unequal size among affected organs, among animals treated alike, and among groups of animals treated differently. We assess the impact of that type of sampling on the precision of the analyzed variable (log of estimated worm burden), derive the variance of the standard nonlinear estimator of efficacy of anthelmintic treatment, and examine the problem of number of animals required for adequate sensitivity of experiments. The standard error of sample geometric mean worm burden, for a particular anthelmintic treatment, and the standard error of estimated efficacy of a treatment, relative to control, are given for the case of log-normal burdens. Small aliquots affect precision critically only if mean burden is small, i.e., when sampling by small aliquots is unnecessary, because the physical effort required is not great. The minimal number of animals per treatment, required for at least 80% power to detect efficacy of .7 or higher, is about 4 to 6 for species of parasites constituting major burdens (where the coefficient of variation of worm burden often is near .7). However, the minimal number of animals may be as high as 15 to 20 per treatment for cases with lowly-abundant species of parasites (where the coefficient of variation may be as high as 2 or 3). An example is given to illustrate procedures.
在统计家畜体内的蠕虫寄生虫（“蠕虫负荷”）时，实际操作上的限制往往导致在受影响的器官之间、接受相同处理的动物之间以及接受不同处理的动物组之间，以大小不等的小份样本进行抽样。我们评估了这种抽样方式对分析变量（估计蠕虫负荷的对数）精度的影响，推导了驱虫治疗效果标准非线性估计量的方差，并研究了实验具有足够灵敏度所需的动物数量问题。对于对数正态分布的负荷情况，给出了特定驱虫治疗的样本几何平均蠕虫负荷的标准误差，以及相对于对照的治疗估计效果的标准误差。只有当平均负荷较小时，小份样本才会严重影响精度，也就是说，当通过小份样本进行抽样没有必要时，因为所需的实际操作难度不大。对于构成主要负荷的寄生虫种类（蠕虫负荷的变异系数通常接近0.7），为了有至少80%的把握检测到0.7或更高的疗效，每种处理所需的最少动物数量约为4至6只。然而，对于寄生虫种类丰富度低的情况（变异系数可能高达2或3），每种处理所需的最少动物数量可能高达15至20只。给出了一个例子来说明具体步骤。
