Multivariate calibration and standardization: From first to second order

The development and application of first order multivariate calibration methods have received sufficient attention in the field of chemometrics in recent years, owing to their ability to accomplish simultaneous multicomponent analysis. The success of first order calibration in real world applications, however, has been largely limited by one of the most difficult problems in process analytical chemistry--recalibration. This dissertation proposes several multivariate standardization methods to avoid recalibration by the measurement of a small common set of transfer samples. Through case studies, this dissertation explores the possibility of utilizing both subset samples and generic standards as transfer samples, demonstrates standardization between different sample temperatures, and investigates improvement of multivariate calibration via standardization.; A precondition for first order calibration is that all components to be encountered in future unknown samples be included in the design of the calibration set. This precondition can be eliminated by the use of second order multivariate calibration, where the instrument response of a sample is recorded as a function of two different variables. Second order calibration allows simultaneous multicomponent quantitation in the presence of unknown interferences. This dissertation focuses on second order calibration with nonbilinear data where the rank of a pure component response is greater than one. The concepts of rank linear additivity, net analyte rank and net analyte signal are developed through theoretical analysis and computer simulation. Other analytical figures of merit, namely, signal to noise ratio, selectivity, sensitivity, and limit of determination can also be defined for second order calibration. It is shown that the net analyte rank for the analyte of interest has to be at least one in order to perform quantitation in the presence of unknown interferences. An example from MS/MS is included for demonstration.; As in first order calibration, second order calibration also suffers from the recalibration problem. A second order standardization method is thus proposed which requires the measurement of only one transfer sample. The proposed method is studied through computer simulation and its applicability is demonstrated on LC/UV experimental data.