A partial-least-squares approach to interpretative analysis of multivariate data

Abstract Kvalheim, O.M., 1988. A partial-least-squares approach to interpretative analysis of multivariate data. Chemometrics and Intelligent Laboratory Systems , 3: 189-197. A partial-least-squares (PLS) approach to the resolution of blocks of variables into interpretable factors is developed. The combination step necessary to obtain such factors from a PLS-decomposed matrix is similar to that of target-transformation of principal components. If the block of ‘dependent’ variables can be resolved into uncorrelated interpretable factors, the performance of the PLS method for interpretative data analysis is improved by using, in the decomposition step, these factors standardized to equal variance. This approach eliminates the effects of differences in relative size of factors in the ‘dependent’ variable block, and thus provides enhanced resolution of small factors which are often masked by the larger ones in truncated principal component (PC) and ordinary PLS decompositions. The ability of the present PLS approach to include small but relevant variation in the early components gives a sharper borderline between noise and information than is obtained with PC and ordinary PLS decompositions.

[1]  H. Hotelling Analysis of a complex of statistical variables into principal components. , 1933 .

[2]  O. Kvalheim,et al.  Visualizing information in multivariate data: Applications to petroleum geochemistry : Part 2. Interpretation and correlation of north sea oils by using three different biomarker fractions , 1986 .

[3]  Oils of the Po Basin: A Geochemical—chemometric Study , 1987 .

[4]  R. L. Mason,et al.  Selecting principal components in regression , 1985 .

[5]  Olav M. Kvalheim,et al.  A general-purpose program for multivariate data analysis , 1987 .

[6]  R. Manne Analysis of two partial-least-squares algorithms for multivariate calibration , 1987 .

[7]  O. M. Kvalheim Interpretation of direct latent-variable projection methods and their aims and use in the analysis of multicomponent spectroscopic and chromatographic data , 1988 .

[8]  I. Helland ON THE STRUCTURE OF PARTIAL LEAST SQUARES REGRESSION , 1988 .

[9]  C. Eckart,et al.  The approximation of one matrix by another of lower rank , 1936 .

[10]  Richard A. Reyment Multivariate analysis in geoscience: Fads, fallacies and the future , 1987 .

[11]  O. Kvalheim,et al.  Visualizing information in multivariate data: Applications to petroleum geochemistry , 1986 .

[12]  S. Wold,et al.  The Collinearity Problem in Linear Regression. The Partial Least Squares (PLS) Approach to Generalized Inverses , 1984 .

[13]  Multivariate calibration : quantitative interpretation of non-selective chemical data , 1985 .

[14]  Terje V. Karstang,et al.  Multivariate calibration of diffuse reflectance infrared spectra of coals as an alternative to rank determination by vitrinite reflectance , 1987 .

[15]  I. Jolliffe A Note on the Use of Principal Components in Regression , 1982 .

[16]  Olav M. Kvalheim,et al.  Latent-structure decompositions (projections) of multivariate data , 1987 .