多元校正分光光度法测定混合色素实验的数据分析

The analysis of multivariate calibration experimental data for mixed pigments measured by vis spectrophotometry

利用纯数学运算的奇异值分解(singularvalue decomposition,SVD)测量矩阵法与具有物理运算意义的K矩阵回归法解析多元校正分光光度法测定混合色素实验数据,对应扣除解析得到的数学主成分与物理主成分后获得了相应的剩余误差信息矩阵E_(数学)和E_(物理)及相应剩余标准偏差(residual standard deviation,RSD)RSD_(数学)和RSD_(物理)值,在此基础上比较了2种剩余矩阵计算方法对于2种RSD值大小关系的影响以及剩余误差的组成和来源;通过对数学剩余矩阵E_(数学)数据特征的直接分析快速判断了测量数据中的单点异常值;基于留一(leave one out,LOO)交叉验证法提出了轮换剔除1个样本交叉验证法用于查找实验误差产生的具体原因。将以上多种化学计量学方法应用到本科实验教学中,一方面加深了学生对化学计量学在实验数据分析过程中应用性的理解,另一方面也拓展了本科实验教学的宽度与深度。

The singular value decomposition（SVD） and K matrix methods were employed to resolve the multivariate calibration experimental data of mixed pigment（Carmine, Sunset yellow, Tartrazine） measured with vis spectrophotometry, by which the mathematic residual matrix of Ermthematlc and Ephssical were obtained and then the residual standard deviation（RSD） of mathematic RSDmathematic and physical RSDphasicat were caculated base on Emathematic and Ephasical . The relationship between two residual matrixes and two RSD values were compared to analyze the origin of experimental errors. The outlier in the eperimental data set was founded by analyzing the Emthematic directly. The alternate excluding one sample cross validation method was put foward base on Leave One Out（LOO） cross validation to find the reasons of error. The above mentioned chemometric methods applied to the undergraduate experimental teaching, on the one hand to deepen the students＇ understanding about the Chemometrics in the application of experimental data analysis process, it also expanded the width and depth of undergraduate experimental teaching.