Facultad de Medicina Veterinaria y Zootecnia, Universidad de Córdoba, Cra 6 # 76-103, Ciudad Universitaria, Montería, Colombia.
Res Vet Sci. 2013 Oct;95(2):405-9. doi: 10.1016/j.rvsc.2013.04.005. Epub 2013 Apr 25.
Nowadays the formula to calculate the sample size for estimate a proportion (as prevalence) is based on the Normal distribution, however it would be based on a Binomial distribution which confidence interval was possible to be calculated using the Wilson Score method. By comparing the two formulae (Normal and Binomial distributions), the variation of the amplitude of the confidence intervals is relevant in the tails and the center of the curves. In order to calculate the needed sample size we have simulated an iterative sampling procedure, which shows an underestimation of the sample size for values of prevalence closed to 0 or 1, and also an overestimation for values closed to 0.5. Attending to these results we proposed an algorithm based on Wilson Score method that provides similar values for the sample size than empirically obtained by simulation.
如今，用于估计比例（如患病率）的样本量计算公式基于正态分布，但也可以基于二项式分布，使用威尔逊得分法可以计算其置信区间。通过比较这两个公式（正态分布和二项式分布），置信区间幅度的变化在曲线的尾部和中心是相关的。为了计算所需的样本量，我们模拟了一个迭代抽样过程，该过程显示对于接近 0 或 1 的患病率值，样本量会被低估，而对于接近 0.5 的患病率值，样本量会被高估。根据这些结果，我们提出了一种基于威尔逊得分法的算法，该算法提供的样本量值与通过模拟经验获得的值相似。
