Use of canonical correlation analysis for the evaluation of chromatographic retention data

Abstract

The retention characteristics of 38 nonionic surfactants having different hydrophobic moieties have been determined by reversed-phase thin-layer chromatography using alumina support and methanol as an organic modifier. The hydrophobicity ( R M values) of each surfactant decreased monotonously with increasing concentration of methanol, the linear correlation between the R M value and the concentration of methanol in the eluent was significant in each instance. Canonical correlation analysis was used to find the relationship between the intercept and slope values (related to the molecular hydrophobicity and specific hydrophobic surface area of surfactants) of the above linear correlation and the molecular substructures of surfactants. To compare the information from canonical correlation analysis and other multivariate mathematical—statistical methods, the same data matrix was evaluated by principal component analysis (PCA) and cluster analysis too. Both relationships in the canonical correlation analysis were significant, proving the significant influence of molecular substructures on the reversed-phase retention behaviour of nonionic surfactants. Significant linear correlations were found between the canonical variates and the first coordinate of the nonlinear map of principal component variables, but no correlation was found between the weighted canonical coefficients and the corresponding PCA parameters suggesting similar but not identical information content of the methods. The distribution of the chromatographic parameters and substructures were identical on the cluster dendrogram and on the two-dimensional nonlinear map of PCA loadings showing the similarity of these methods.