Gargano F, Gullotti A, Brancato P, Campione G, Acquaro D
Quad Sclavo Diagn. 1984 Dec;20(4):399-417.
We chose, as typical values of serum cholesterol, the mean levels of the unimodal distribution of frequency with asymmetrical form, for different age groups, from some quarters of Palermo. We assumed as normal range estimate the interval between the 2.5 and 97.5 percentiles, with associated nonparametric 90% confidence intervals for normal limits for the true percentile. For this biometric analysis of ethnic physiology, we controlled the laboratory measurements in several inspection cycles. As the observed frequencies are not a real statistical sample, neither the 95% of the actual reference population: we smoothed the observed frequencies interpolating, with an algebraic polynomial, the mean empiric levels (and the chosen percentiles) of total serum cholesterol of the examined population - since the measurements taken on Mondays, the days after holidays, in January and April were chosen to see, experimentally, how the collected data depend on environmental alimentary factors; the degree of polynomial which gave the best approximation was chosen, on the grounds of the variance analysis; we tested the goodness of fit of the function to the referred frequencies. Moreover, the choice of the function was based on: a) the test of the hypothesis that the dependent variable reading the various specific ages, is normal, so we estimated the kurtosis and the asymmetry of the single frequency distributions; b) the simulation of the hypothesis that the measurements were made at random from gaussian distribution, so that the variances homogeneity of the examined character at various specific ages was examined by Bartlett's test. To identify the stages within which the cholesterolemic levels vary, due to growth and to the sexual maturation, we formed three age groups for males and four for females, whose delimitation is not arbitrary, since the best interpolation of the mean levels observed within such delimited age groups was again obtained by an algebraic polynomial of 2nd degree.
我们选取了巴勒莫部分地区不同年龄组不对称形式单峰频率分布的血清胆固醇平均水平，作为血清胆固醇的典型值。我们将2.5百分位数和97.5百分位数之间的区间作为正常范围估计值，并给出了真实百分位数正常范围的相关非参数90%置信区间。对于这项种族生理学的生物统计学分析，我们在几个检测周期中对实验室测量进行了控制。由于观察到的频率并非真正的统计样本，实际参考人群的95%也不是：我们用代数多项式对观察到的频率进行插值平滑，得到所检查人群总血清胆固醇的平均经验水平（以及选定的百分位数）——因为我们选择了周一、节假日后的日子、1月和4月进行测量，以便通过实验观察收集的数据如何依赖于环境饮食因素；根据方差分析选择能给出最佳近似值的多项式次数；我们检验了该函数与参考频率的拟合优度。此外，函数的选择基于：a）检验假设，即读取不同特定年龄的因变量呈正态分布，因此我们估计了单个频率分布的峰度和偏度；b）模拟假设，即测量是从高斯分布中随机进行的，因此通过Bartlett检验检查了所检查特征在不同特定年龄的方差齐性。为了确定由于生长和性成熟导致胆固醇水平变化的阶段，我们将男性分为三个年龄组，女性分为四个年龄组，其划分并非随意确定，因为在这些划定的年龄组内观察到的平均水平的最佳插值再次通过二次代数多项式获得。
