Ilijević Konstantin, Obradović Marko, Jevremović Vesna, Gržetić Ivan
Faculty of Chemistry, University of Belgrade, Studentski Trg 12-16, 11000, Belgrade, Serbia,
Environ Monit Assess. 2015 Sep;187(9):553. doi: 10.1007/s10661-015-4740-y. Epub 2015 Aug 5.
We have assembled and assessed the statistical procedure which is capable to objectively explore influence of the Danube's major tributaries (the Rivers Tisa, Sava, and Velika Morava) to its eco-chemical status. Procedure contains several tests for measurement of central tendencies: one-way analysis of variance (ANOVA), repeated measures ANOVA, and nonparametric Kruskal-Wallis and Mann-Whitney tests. Various nuisance factors, (outliers, departures from normality, seasonality, and heteroscedasticity) which are present in large data bases, affect the objectivity of central tendency tests; therefore, it was important not only to estimate their robustness, but also to apply proper procedures for detection of the nuisance factors (Grubbs', generalized ESD-extreme Studentized deviate, Kolmogorov-Smirnov, Shapiro-Wilk, turning point, Wald-Wolfowitz runs, Kendall rank, and Levene's tests) and to mitigate their influence (outlier exclusion, Box-Cox, and logarithmic transformations). The analysis of selected eco-chemical parameters: biological oxygen demand-5, chemical oxygen demand, UV extinction at 254 nm, dissolved oxygen, oxygen saturation, total dissolved solids, electrical conductivity, suspended matter, total phosphorus, phosphates, nitrates, ammonia, pH, total alkalinity, m-2p alkalinity, CO2, and temperature, was performed for 15 years period. The Tisa was the most polluted tributary, but its pollution load was not substantial enough to exceed the Danube self-purification potential. The City of Belgrade was also identified as serious pollution source. Assessment of assembled statistical procedure, which was based on the real environmental data, indicates that proposed tests are sufficiently robust to the observed level of nuisance factors with the exception of pronounced seasonality.
我们已经汇总并评估了一种统计程序，该程序能够客观地探究多瑙河的主要支流（蒂萨河、萨瓦河和大摩拉瓦河）对其生态化学状况的影响。该程序包含多个用于测量集中趋势的测试：单因素方差分析（ANOVA）、重复测量方差分析以及非参数克鲁斯卡尔 - 沃利斯检验和曼 - 惠特尼检验。大型数据库中存在的各种干扰因素（异常值、偏离正态性、季节性和异方差性）会影响集中趋势测试的客观性；因此，不仅要估计这些测试的稳健性，还要应用适当的程序来检测干扰因素（格拉布斯检验、广义ESD - 极端学生化偏差检验、柯尔莫哥洛夫 - 斯米尔诺夫检验、夏皮罗 - 威尔克检验、转折点检验、沃尔德 - 沃尔福威茨游程检验、肯德尔秩检验和莱文检验）并减轻它们的影响（异常值排除、Box - Cox变换和对数变换）。对选定的生态化学参数进行了为期15年的分析，这些参数包括：五日生化需氧量、化学需氧量、254纳米处的紫外线吸光度、溶解氧、氧饱和度、总溶解固体、电导率、悬浮物、总磷、磷酸盐、硝酸盐、氨、pH值、总碱度、m - 2p碱度、二氧化碳和温度。蒂萨河是污染最严重的支流，但其污染负荷不足以超过多瑙河的自净潜力。贝尔格莱德市也被确定为严重污染源。基于实际环境数据对汇总的统计程序进行的评估表明，除了明显的季节性外，所提出的测试对观察到的干扰因素水平具有足够的稳健性。
