论文部分内容阅读
用修正的ｊａｃｋｋｎｉｆｅ检验研究了估算有机化合物吸着系数（Ｋｏｃ）的分子拓扑－极性校正模型及模型参数的稳健性。检验分四种不同方式进行：随机抽除单一化合物（１００次）、逐一抽除异常化合物（估算值与实测值差别在０．９个对数单位以上者，２７次）、随机抽除５０个化合物（３０次）以及逐一抽除特定化合物类别（１７类）。检验结果表明，对所研究的模型，不同检验方式得到的结果一般具有相似的趋势，即多元回归模型的可决系数表现出较强的统计稳健性，相比之下，回归系数的稳健性较低，特别是六阶链指数６Ｘｃｈ。在１７种极性校正因子中，ＳＯ２、 Ｏ、 Ｆ．ＮＨ２和 ＳＯ等的稳健性相对较低。
The modified molecular jackknife test was used to study the molecular topological-polarity correction model and the robustness of the model parameters to estimate the Koc adsorption coefficient (Koc). The tests were performed in four different ways: random compound (100 times) was withdrawn and abnormal compounds were removed one by one (the difference between the estimated value and the measured value was more than 0.9 logarithm units, 27 times) Compounds (30 times) and one by one compound-specific categories (17 categories). The test results show that the results obtained by different test methods tend to have similar trends for the studied models, that is, the coefficient of the multiple regression model shows a strong statistical robustness, compared with the robustness of the regression coefficients Low, especially the sixth-order chain index 6Xch. Among the 17 kinds of polarity correction factors, SO2, O, and F. The robustness of NH2 and SO is relatively low.