Journal of IiME Volume 1 Issue 2 www.investinme.org Visible and near-infrared (Vis-NIR) spectroscopy: Introduction and Perspectives for Diagnosis of CFS HIV: human immunodeficiency virus principal component analysis PCA: PLS: Table 1. Representative biomolecules studied by Vis-NIR spectroscopy MLR: multiple linear regression analysis PCRA: principal component regression analysis partial least squares regression analysis Vis-NIR spectroscopy is not very sensitive: the limit is only about 0.15% (w/w) for most constituents, and the signal to noise ratio of the instrument is low [less than 10-4 optical density (OD)] [76], but is dependent on several factors such as the measurement accessory, spectrometer including detectors, and acquisition time. A large amount of sample is needed for Vis-NIR spectroscopy compared to other methods of chemical analysis [76]. The direct interpretation of spectral absorbance is very difficult for complex mixtures because of broad overlapping and interacting absorption bands [76]. Vis-NIR spectroscopy thus relies on a multivariate analysis to quantify properties or constituents of interest. A multivariate analysis is an analysis of data with many variables based on statistics and mathematics. It can simplify complicated data and uncover hidden information. The analysis can be qualitative or quantitative. It is based on chemometrics algorithms. Methods of quantitative analysis include the partial least Invest in ME Charity Nr 1114035 SIMCA: software-independent modeling by class analogy squares regression analysis (PLS) and the principal component regression analysis (PCRA), which are used to develop the regression model for the prediction of the reference value [77, 78]. Methods of qualitative analysis include the principal component analysis (PCA) [79] and the software-independent modeling by class analogy (SIMCA) [80]. PCA is a method for transforming an original variable such as absorbance at various wavelengths into new variables called principal components (PCs). By plotting the data defined by PCs, important relationships in the data (e.g., similarities and differences among objects) can be clearly identified. SIMCA is a recently developed method based on PCA [81]. PCA reduces the amount of data, and SIMCA further extracts discriminant rules among different groups. PCRA is a method for performing PCA on x variables such as wavelength and then regressing y variables on the principal components, whereas PLS (continued on page 11) Page 10/72

11 Publizr Home

You need flash player to view this online publication