Abstract
A new multivariate calibration technique called net analyte preprocessing is introduced for the study of mixtures by near infrared spectroscopy. It involves the construction of a space spanned by all sources of variability except one component (or reference property). Both the data matrix and the unknown sample spectrum are preprocessed through a projection orthogonal to the latter space, yielding a set of net analyte signals appropriate for quantifying a particular mixture component. The construction of the net analyte space is carried out using a certain number of spectral factors, which can be estimated by cross-validation. Advantages of the presently discussed method include the estimation of figures of merit (sensitivity, selectivity and limit of determination) and the presentation of data in a useful pseudounivariate mode. An example involving the successful analysis of mixtures of the two rubber antioxidants tris-(nonylphenyl)-phosphite and 4,6-bis-(octylthiomethyl)-o-cresol is described, and compared with classical techniques such as principal component regression and partial least-squares.
© 2001 NIR Publications
PDF Article
More Like This
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription