Principal components regression for routine multicomponent UV determinations: A validation protocol

Abstract

SUMMARYA validation protocol for multicomponent spectroscopic assays based on principal components regression is described. Factorial design and hypothesis tests are used to establish the linearity and absence of interaction between components in the regression model. Testing considers multiple response variables simultaneously so that correlation between residuals is properly treated. Assay reproducibility and sensitivity to related substances are evaluated.

Nonlinear multivariate calibration using principal components regression and artificial neural networks

“…Six different calibration algorithms were tried on each data set: (1) principal component regression (PCR), (2) partial least squares (PLS), (3) weighted PCR (WPCR), (4) quadratic PCR (QPCR), (5) artificial neural networks (ANN), and (6) artificial neural networks with orthogonal principal components (PC-ANN). The performance of each method is shown in Tables I11 through V. The simulated two-component and threecomponent linear data sets serve as benchmarks for judging the performance of nonlinear calibration algorithms on the nonlinear data sets.…”

“…We quickly encountered problems with some of the assays, which we attributed to nonlinear spectral response (1). The symptoms included a lack of short-term and long-term reproducibility (large bias between days, ca.…”

Key wavelengths screening using competitive adaptive reweighted sampling method for multivariate calibration

“…But the measured spectral data on the modern spectroscopic instrument, such as ultraviolet or near infrared instruments, are usually of high colinearity, which is the commonplace faced by analytical chemists. To address this problem, a variety of techniques based on latent variables (LVs) have been proposed, such as principal component regression (PCR) [1,2] and partial least squares (PLS) [3,4]. Typically, the establishment of a calibration model usually includes all the measured wavelengths.…”

Application of Near-Infrared Spectroscopy for Non-Invasive Determination of Blood/Tissue Glucose Using Neural Networks

“…In real world problems the assumptions which lead to the application of the linear model are often violated leading to real and apparent non-linear spectral response, which can be attributed to instrumental, physical and chemical sources, e.g. curvature in the concentration response function and shifts in the position of absorption bands [16,17]. In these cases, artificial neural networks (ANN) as a model-free approach have been used with success and surpassed other methods.…”