论文部分内容阅读
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 pure math operation singular value decomposition (SVD) measurement matrix method and the physical operation of the K matrix regression analysis of multivariate calibration spectrophotometric determination of mixed pigment experimental data, corresponding deduction analysis of the mathematical principal components and the physical host We obtain the corresponding RSD (math) and RSD (physical) values of E_ (math) and E_ (physical) and corresponding residual standard deviation (RSD) The effect of residual matrix calculation method on the relationship between the two kinds of RSD values and the composition and source of the residual error; through the direct analysis of E_ (math) data features of math residual matrix, the single point anomaly in measurement data can be quickly judged; (leave one out, LOO) cross-validation method proposed a rotation elimination of a sample cross-validation method for finding the specific causes of experimental error. Applying the above various chemometrics methods to undergraduate experiment teaching, on the one hand, students’ understanding of the applicability of chemometrics in experimental data analysis is deepened, on the other hand, the breadth and depth of undergraduate experiment teaching are also broadened.