PCA-ContVarDia: an improvement of the PCA-VarDia technique for curve resolution in GC-MS and TG-MS analysis

Principal component analysis (PCA) and variance diagram (VarDia) technique have been used for curve resolution in time-resolved mass spectrometry data . The VarDia shows the clustering of the mass variables in a two-dimensiona l (2D) principal component (PC) subspace. A cluster in the VarDia is an ind ication of the direction of a component axis. However, in most cases a 2D P C subspace cannot provide simultaneously, information for all the component axes. In this work, an improvement of the VarDia is presented which aims t o the faster and easier determination of more components in an unresolved t otal ion current curve. In this improvement the loadings of the mass variab les are plotted in a three-dimensional (3D) PC subspace. This subspace is s canned in steps, systematically, using spherical coordinates and 3D "window s". The variance of the mass vectors present in every 3D "window" is calcul ated in steps. The contour variance diagram (ContVarDia) which is a contour plot is used for the visualization of the calculated variance versus the s pherical coordinates. An area of high variance in the ContVarDia is an indi cation of the direction of a component axis. A set of simulated data corres ponding to an unresolved gas chromatography-mass spectrometry (GC-MS) or th ermogravimetry-mass spectrometry (TG-MS) peak and a set of real unresolved GC-MS data consisted of four compounds were used for evaluating the PCA-Con tVarDia method. The results of the application of the PCA-ContVarDia method are presented and discussed. (C) 2001 Elsevier Science B.V. All rights res erved.