Soliman S A, Abdel-Hay M H, Tayeb O S, Sulaiman M I
Clin Chim Acta. 1987 Jun 30;166(1):9-16. doi: 10.1016/0009-8981(87)90189-6.
We have applied a multi-stage computer algorithm for normalization of distributions and calculation of reference intervals of some urine characteristics and constituents. The study analyzed 24-h urines, collected from adult male Saudis from different socioeconomic classes, for volume, pH, osmolality, specific gravity, creatine, creatinine, urea and uric acid. Frequency distributions, for each analyte, were found to be nongaussian as judged by the coefficients of skewness and kurtosis, chi 2 and Kolmogorov-Smirnov tests, and from probability plots. Data were transformed to gaussian distributions by multistage log-power transformation. Stepwise, this procedure removed skewness and residual kurtosis. Using the gaussian transformed data the reference intervals were estimated parametrically as the mean +/- 2 SD. In addition, the non-parametric percentile technique was applied to estimate these values. The former intervals were found to have narrower 0.90 confidence limits than the latter. When established limits were compared with those reported for Western subjects urine volume and uric acid showed the most marked variation.
我们应用了一种多阶段计算机算法来对某些尿液特征和成分的分布进行标准化，并计算其参考区间。该研究分析了从不同社会经济阶层的成年沙特男性收集的24小时尿液，检测了尿量、pH值、渗透压、比重、肌酸、肌酐、尿素和尿酸。通过偏度系数、峰度系数、卡方检验和柯尔莫哥洛夫-斯米尔诺夫检验以及概率图判断，发现每种分析物的频率分布均为非高斯分布。通过多阶段对数幂变换将数据转换为高斯分布。逐步地，该过程消除了偏度和残余峰度。使用高斯变换后的数据，通过参数估计将参考区间估计为均值±2标准差。此外，还应用了非参数百分位数技术来估计这些值。发现前者的0.90置信限比后者更窄。当将既定的限值与针对西方受试者报告的限值进行比较时，尿量和尿酸显示出最明显的差异。
