Comparison of two principal component analysis methods to evaluate reversed-phase retention data

Tibor Cserháti*, Zoltán Illés

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract (may include machine translation)

The retention of twelve 2-nitro-4-cyanophenyl esters showing marked herbicidal activity was determined in 23 reversed-phase thin-layer chromatographic systems. The retention data set was evaluated by principal component analysis (PCA). To assess the effect of the information loss caused by normalization, PCA was separately carried out on the covariance (method A) and on the correlation matrix (method B). The ratio of the variances explained was very similar for both methods, however, the PC loadings and the coordinates of the two-dimensional nonlinear maps showed poor correlation. The distribution of the 2-nitro-4-cyanophenyl esters and that of chromatographic systems showed differences on the two-dimensional nonlinear maps of PC loadings and PC variables, however, the general trend was similar independently of the application of method A or B. The findings indicate that the application of the correlation matrix as basis for the PCA calculations may lead to slightly distorted results that strongly advocates the use of covariance matrix in PCA.

Original languageEnglish
Pages (from-to)685-691
Number of pages7
JournalJournal of Pharmaceutical and Biomedical Analysis
Volume9
Issue number9
DOIs
StatePublished - 1991
Externally publishedYes

Keywords

  • Principal component analysis
  • correlation matrix
  • covariance matrix
  • stepwise regression analysis.
  • two-dimensional nonlinear mapping

Fingerprint

Dive into the research topics of 'Comparison of two principal component analysis methods to evaluate reversed-phase retention data'. Together they form a unique fingerprint.

Cite this