【摘要】 主成分分析是全光谱分光光度分析中常用的校正方法。本文提出第一主成分并不是与因变量最线性相关的主成分.为此,我们利用扫描算法从众多主成分中选择与因变量（浓度）最相关的主成分,从而使计算结果更准确可信.本文还对单因变量和多因变量两种情况下主成分选择的统计量进行了讨论.更多还原

【Abstract】 Principal component analysis is widely applied to the multivariate calibration. In principal component-spectrophotometry, the first several principal components regress with concentration to get regression coffecient. But the first principal component may not be a best linear correlated with concentration. We use scan algorithms method for the choice of several principal components that are best linear correlated with concentration from a lot of principal components. These principal components regress with concentrations to get regression coffecients. These regression coffecients are applied to the prediction of concentration of unknown sample. A program written in Turbo BASIC has been applied to the quantitative analysis of the Fourier transform near infrared diffuse reflectance spectroscopy of wheat sample and UV spectroscopy of six amino acids mixture with satisfactory results.更多还原