摘要
用MEDV描述子对102个酚的分子结构进行表征,对MEDV矢量和其气相色谱保留指数之间建立了分子结构-色谱保留值定量关系(QSRR)。多元线性回归十参数模型的复相关系数达到了0.9436,逐步回归九参数模型复相关系数为0.9436,交互检验的RCV值为0.8448,表明模型对样本具有一定的稳定性和预测能力。
The molecular electronegativity-distance vector (MEDV)descriptors were used to characterize the chemical structure of 102 phenols. The MEDV descriptors were employed to establish the quantitative relationship models with respect to retention indices of gas chromatography (RIGC) for 102 compounds examined. The correlation coefficient of the ten-variables models obtained from multiple linear regression (MLR)is R=0.943 6, and that of the nine-variables model obtained from stepwise multiple regression (SMR)is R=0.943 6, the correlation coefficient value of in cross-validation by the leave-one-out procedure (LOO)is R_(CV)=0.844 8. The satisfactory results showed the constructed models were provided with excellent estimation stability and favorable predictive ability.
出处
《重庆大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2004年第4期67-71,共5页
Journal of Chongqing University
基金
国家"春晖计划"教育部启动基金资助项目(99)
重庆市应用基础项目资助课题(01)