论文部分内容阅读
用检测器阵列进行多组分定量分析时,体系常不同程度地受复共线关系的影响,致使多种基于最小二乘原理的校正方法难于获得稳定的预测结果.为此,本文在文献的基础上,提出了对每一组分分别建立预测模型的LRE(latent Root Estimator)方法.5组分药物体系的定量分析表明本法具有良好的实用性,其浓度预测结果优于最小二乘及主成分估计.
In the multi-component quantitative analysis with detector arrays, the systems are often affected by the complex collinearity to some extent, which makes it difficult to obtain a stable prediction result by many kinds of correction methods based on the least-squares principle.Therefore, On the basis of this, we propose a latent root estimator (LRE) method to establish a predictive model for each component.Quantitative analysis of the 5-component drug system shows that this method has good practicability and its concentration prediction is better than the least squares and Principal component estimation.