Usui T
Department of Laboratory Medicine, Nagasaki University School of Medicine.
Rinsho Byori. 1992 Aug;40(8):820-7.
There are many methods for normal range calculation, and they give different values even with the same normal data sample. I proposed three parameters, accuracy, precision and robustness, for evaluating the calculation methods. A complicated calculation method, the distribution selection method, seems to be theoretically correct, but it shows unstable precision and inadequate robustness when the normal data contain a small amount of abnormal data. A method called "Clinical reference range program (CRRP)" is a robust method, and is applicable to the data which contains abnormal data. I also proposed a condition which defines the sample size required for the calculation using the precision parameter.
正常范围的计算方法有很多种，即使对于相同的正常数据样本，它们给出的值也不同。我提出了三个参数，即准确性、精密度和稳健性，用于评估计算方法。一种复杂的计算方法，即分布选择法，在理论上似乎是正确的，但当正常数据包含少量异常数据时，它显示出不稳定的精密度和不足的稳健性。一种称为“临床参考范围程序（CRRP）”的方法是一种稳健的方法，适用于包含异常数据的数据。我还提出了一个条件，该条件使用精密度参数定义计算所需的样本量。
