摘要
用分子电距矢量描述子(MEDV)对24个从慈竹叶中分离出的挥发油成分的分子结构进行了表征,应用多元线性回归技术建立了结构描述子MEDV与这些化合物的色谱保留时间的定量相关模型。对9参数模型进行线性回归,复相关系数达到了0 995,但交互检验的相关系数为-3 019。对此模型进行改进,采用逐步回归5参数模型,复相关系数为0.993,交互检验的相关系数RCV值提高到0 984,表明5参数模型对样本具有良好的估计和预测能力。
The molecular electrongativity-distance vector(MEDV) descriptors are used to express the structures of 24 chemical compounds isolated from the volatile oil of bamboo leaves.Using multiple linear regression(MLR) technique,the quantitative relationship model is established for the MEDV descriptors and the retention times of gas chromatography.The correlation coefficient of the 9-variable model obtained from MLR is 0.995,however,the value of cross validations is R_(CV)=–3.019.By improving the model,the correlation coefficient of the 5-variable model obtained from stepwise multiple regression (SMR) is 0.993,and the increased value of R_(CV)is 0.984.The results show that the constructed model is provided with good estimation stability and favorable predictive ability.
出处
《精细化工》
EI
CAS
CSCD
北大核心
2004年第11期815-818,836,共5页
Fine Chemicals
基金
重庆应用基础[2001]
国家"春晖计划"教育部启动基金[1999]
霍英东基金[1998]
重庆大学创新基金[03-5-6]资助项目~~
关键词
分子电距矢量
挥发油
色谱保留时间
molecular electronegativity-distance vector
volatile oil
retention time of gas chromatography
