The following is an abstract for the selected article. A PDF download of the full text of this article is available here. Members may download full texts at no charge. Non-members may be charged a small fee for certain articles.

Carbon Number Prediction from Herschel-Infrared Spectra Using Partial Least-Squares Regression

Volume 45, Number 4 (May 1991) Page 713-714

Schrieve, Garin D.; Ullman, Alan H.

In the accompanying note, we pointed out the potential usefulness of the Herschel-infrared (∼700-1100 nm) for solvent measurements, particularly for process measurements over optical fibers and with filter photometers. In this note we demonstrate that multivariate mathematics can be used to extract even more information, which would be difficult or impossible to obtain directly from the spectra. The subtle differences in the spectra of a homologous series of n-alkanes allowed us to use partial least-squares regression (PLS) to model and predict the carbon chain length of the alkanes.