The Use of Principal Components in the Analysis of Near-Infrared Spectra

Cowe, Ian A.; McNicol, James W.

doi:10.1366/0003702854248944

Cited by 264 publications

(110 citation statements)

References 7 publications

Supporting

Mentioning

107

Contrasting

Unclassified

Order By: Relevance

“…Later, this approach was largely superseded by methods that reduce the p spectral variables to scores on a much smaller number of factors and then regress on these scores. Two variants-principal components regression (PCR; Cowe and McNicol 1985) and partial least squares regression (PLS; Wold, Martens, and Wold 1983)-are now widely used, with equal effectiveness, as the standard approaches. The increasing power of computers has triggered renewed research interest in wavelength selection, now using computer-intensive search methods.…”

Section: Standard Analysesmentioning

confidence: 99%

Bayesian Wavelet Regression on Curves With Application to a Spectroscopic Calibration Problem

Brown

Fearn

Vannucci

2001

Journal of the American Statistical Association

157

117

View full text Add to dashboard Cite

Motivated by calibration problems in near-infrared (N IR) spectroscopy, we consider the linear regression setting in which the many predictor variables arise from sampling an essentially continuous curve at equally spaced points and there may be multiple predictands. We tackle this regression problem by calculating the wavelet transforms of the discretized curves, then applying a Bayesian variable selection method using mixture priors to the multivariate regression of predictands on wavelet coef cients. For prediction purposes, we average over a set of likely models. Applied to a particular problem in N IR spectroscopy, this approach was able to nd subsets of the wavelet coef cients with overall better predictive performance than the more usual approaches. In the application, the available predictors are measurements of the N IR re ectance spectrum of biscuit dough pieces at 256 equally spaced wavelengths. The aim is to predict the composition (i.e., the fat, our, sugar, and water content) of the dough pieces using the spectral variables. Thus we have a multivariate regression of four predictands on 256 predictors with quite high intercorrelation among the predictors. A training set of 39 samples is available to t this regression. Applying a wavelet transform replaces the 256 measurements on each spectrum with 256 wavelet coef cients that carry the same information. The variable selection method could use subsets of these coef cients that gave good predictions for all four compositional variables on a separate test set of samples. Selecting in the wavelet domain rather than from the original spectral variables is appealing in this application, because a single wavelet coef cient can carry information from a band of wavelengths in the original spectrum. This band can be narrow or wide, depending on the scale of the wavelet selected.

show abstract

Section: Standard Analysesmentioning

confidence: 99%

Bayesian Wavelet Regression on Curves With Application to a Spectroscopic Calibration Problem

Brown

Fearn

Vannucci

2001

Journal of the American Statistical Association

157

117

View full text Add to dashboard Cite

show abstract

“…Principle component analysis (PCA) [11] was used to describe multidimensional patterns of the ET dataset and to discover outliers. Quantitative models were built using partial least squares regression (PLSR) [12], multiple linear regression (MLR) [13] and support vector machine regression (SVM) [14] for the prediction of the physicochemical properties using the data of electronic tongue, pH and EC.…”

Section: Discussionmentioning

confidence: 99%

Prediction of main analytical and physical parameters of honey with electronic tongue

Koncz¹,

Bodor²,

Kaszab³

et al. 2017

HAE

View full text Add to dashboard Cite

Honey is one of the most commonly adulterated product in the food market. The different types of adulterations affect the market negatively, therefore an effective honey evaluation method is required. The electronic tongue could be a new alternative tool for inspection. In this study 78 authentic Hungarian honey samples were analysed with electronic tongue. The main analytical and physical parameters of honey samples were also determined, with classical analytical methods. Multivariate regression models (PLSR, MLR, SVM) were built to predict the main physicochemical properties of honey based on the results of electronic tongue. Results showed that the merged data of electronic tongue and electrical conductivity provided the best models for the prediction of main physicochemical properties of honey.

show abstract

“…Principal component analysis (PCA) was applied to describe multidimensional patterns of the NIR dataset and to discover outliers [20]. Principal component regression (PCR) and partial least squares regression (PLSR) were used for the calibration models on individual blood clinicochemical parameters [15].…”

Section: Data Evaluationmentioning

confidence: 99%

Multicomponent blood lipid analysis by means of near infrared spectroscopy, in geese

Bázár

Éles

Kovács

et al. 2016

Talanta

View full text Add to dashboard Cite

a b s t r a c tThis study provides accurate near infrared (NIR) spectroscopic models on some laboratory determined clinicochemical parameters (i.e. total lipid (5.57 71.95 g/l), triglyceride (2.59 71.36 mmol/l), total cholesterol (3.81 7 0.68 mmol/l), high density lipoprotein (HDL) cholesterol (2.45 70.58 mmol/l)) of blood serum samples of fattened geese. To increase the performance of multivariate chemometrics, samples significantly deviating from the regression models implying laboratory error were excluded from the final calibration datasets. Reference data of excluded samples having outlier spectra in principal component analysis were not marked as false. Samples deviating from the regression models but having non outlier spectra in PCA were identified as having false reference constituent values. Based on the NIR selection methods, 5% of the reference measurement data were rated as doubtful. The achieved models reached R 2 of 0.864, 0.966, 0.850, 0.793, and RMSE of 0.639 g/l, 0.232 mmol/l, 0.210 mmol/l, 0.241 mmol/l for total lipid, triglyceride, total cholesterol and HDL cholesterol, respectively, during independent validation. Classical analytical techniques focus on single constituents and often require chemicals, timeconsuming measurements, and experienced technicians. NIR technique provides a quick, cost effective, non-hazardous alternative method for analysis of several constituents based on one single spectrum of each sample, and it also offers the possibility for looking at the laboratory reference data critically. Evaluation of reference data to identify and exclude falsely analyzed samples can provide warning feedback to the reference laboratory, especially in the case of analyses where laboratory methods are not perfectly suited to the subjected material and there is an increased chance of laboratory error.

show abstract

The Use of Principal Components in the Analysis of Near-Infrared Spectra

Cited by 264 publications

References 7 publications

Bayesian Wavelet Regression on Curves With Application to a Spectroscopic Calibration Problem

Bayesian Wavelet Regression on Curves With Application to a Spectroscopic Calibration Problem

Prediction of main analytical and physical parameters of honey with electronic tongue

Multicomponent blood lipid analysis by means of near infrared spectroscopy, in geese

Contact Info

Product

Resources

About