The Collinearity Problem in Linear Regression. The Partial Least Squares (PLS) Approach to Generalized Inverses

Wold, Svante; Ruhe, Axel; Wold, Herman; Dunn, William J.

doi:10.1137/0905052

Cited by 2,188 publications

(1,379 citation statements)

References 30 publications

Supporting

Mentioning

1,360

Contrasting

Unclassified

Order By: Relevance

“…A supervised partial least squares discriminant analysis (PLS-DA) was performed to maximize the separation between different groups of the gill tissue samples (Xu, 2004;Wold et al, 1984). The PLS-DA included a measured matrix X and a response matrix Y consisting of dummy variables.…”

Section: Spectral Data Pre-processing and Multivariate Analysismentioning

confidence: 99%

Effects of hypoxia in the gills of the Manila clam Ruditapes philippinarum using NMR-based metabolomics

Zhang

Wei

et al. 2017

Marine Pollution Bulletin

View full text Add to dashboard Cite

Coastal hypoxia affects the survival, behavior, and reproduction of individual local marine organisms, and the abundance, biomass, and biodiversity of coastal ecosystems. In this study, we investigated the chronic effects of hypoxia on the metabolomics in the gills of Ruditapes (R.) philippinarum. The results indicated significant alterations in the metabolite profiles in the gills of the hypoxia-treated clams, in comparison with those maintained under normoxia. The levels of betaine, taurine, glycine, isoleucine, and alanine were significantly reduced, suggesting a disturbance of osmotic balance associated with hypoxia. Meanwhile, metabolites involved in energy metabolism, such as alanine and succinate, were also affected. Dramatic histopathological changes were observed in the gills and hepatopancreases of R. philippinarum grown in hypoxic waters, demonstrating tissue damages apparently caused by long-term exposure to hypoxia. Our findings suggest that hypoxia significantly affects the physiology of R. philippinarum, even at a sub-lethal level, and impedes health of the clams.

show abstract

Section: Spectral Data Pre-processing and Multivariate Analysismentioning

confidence: 99%

Effects of hypoxia in the gills of the Manila clam Ruditapes philippinarum using NMR-based metabolomics

Zhang

Wei

et al. 2017

Marine Pollution Bulletin

View full text Add to dashboard Cite

show abstract

“…Both LOO and LNO CV have been implemented the advantages and disadvantages of these two approaches are discussed below. The SAMPLS algorithm [13] provides a highly efficient implementation of PLS-1 and, for univariate Y only, gives identical results to the classical NIPALS [5] and SIMPLS [14] algorithms.…”

Section: Searching For An Optimal Solution (V Opt ) Eva_gamentioning

confidence: 99%

“…The number of variables is thus very much larger than the number of compounds in a standard QSAR dataset and the Partial least squares to Latent Structures (PLS) technique [5] has been used to provide a robust regression analysis. The purpose of the EVA smoothing procedure is not to simulate an experimental IR spectrum (transition dipole data is not used and, therefore, all kernels are of fixed maximum amplitude) but rather it is to apply a density function such that vibrations at slightly different frequencies in different compounds can be "overlapped"…”

Section: Classical Evamentioning

confidence: 99%

“…At the same time there is a requirement to search for an optimal set of these σ values and, for reasons explained below, a genetic algorithm (GA) has been used to direct this search. PLS [5] crossvalidation regression scores are used as the fitness function to be optimised by the GA. The proposed technique is fundamentally different from more standard variable selection techniques [6][7][8] in as much as variables are not included/excluded by the procedure; rather, it is their information content (i.e., the selection of frequencies contributing to a variable) that is altered through the adjustment of kernel overlap.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Untitled

Turner

Willett

2000

Journal of Computer-Aided Molecular Design

View full text Add to dashboard Cite

SummaryThe EVA structural descriptor, based upon calculated fundamental molecular vibrational frequencies, has proved to be an effective descriptor for both QSAR and database similarity calculations. The descriptor is sensitive to 3D structure but has an advantage over fieldbased 3D-QSAR methods inasmuch as structural superposition is not required. The original technique involves a standardisation method wherein uniform Gaussians of fixed standard deviation (σ) are used to smear out frequencies projected onto a linear scale. This smearing function permits the overlap of proximal frequencies and thence the extraction of a fixed dimensional descriptor regardless of the number and precise values of the frequencies. It is proposed here that there exist optimal localised values of σ in different spectral regions; that is, the overlap of frequencies using uniform Gaussians may, at certain points in the spectrum, either be insufficient to pick up relationships where they exist or mix up information to such an extent that significant correlations are obscured by noise. A genetic algorithm is used to search for optimal localised σ values using crossvalidated PLS regression scores as the fitness score to be optimised. The resultant models are then validated against a previously unseen test set of compounds. The performance of EVA_GA is compared to that of EVA and analogous CoMFA studies.

show abstract

“…performed using the partial least squares method developed by Wold and coworkers (Wold et al 1984) with cross-validation to give some measure of the predictive ability of the potential energy at each grid point. The result can then be expressed as a 3-dimensional contour surface reflecting the relationship between the molecular field and a specific biological activity.…”

mentioning

confidence: 99%