A simulation study was conducted to evaluate the feasibility of using chemometrics methods to analyze process nuclear magnetic resonance (NMR) data. Using the computer-generated NMR data, training sets and validation sets were constructed to represent several real-world application scenarios. The experimental factors (the spectral noise, the reference measurement error, and the nonlinearity) that affect the performance of a partial least-square (PLS) model were systematically investigated.
Xiaonian Lai, Maziar Sardashti, Bobby J. Lane, Jason J. Gislason, and Daniel J. O'Donnell, "Simulation Studies on Modeling Process NMR Data by Using Chemometrics Approach," Appl. Spectrosc. 54, 54-61 (2000)