Abstract
The proliferation of applications for near infrared (NIR) spectroscopy has been fostered by advances in instrumentation and statistics. NIR analytical instrumentation is becoming more stable and reliable. Chemometrics is playing an important role in qualitative and quantitative NIR spectra analysis. The objective of this study was to evaluate the performances of four commonly used calibration models: (1) stepwise multiple linear regression (SMLR); (2) classical least-squares (CLS); (3) principal component regression (PCR); and (4) partial least-squares (PLS) in NIR spectroscopy analysis when random noise is present in the optical data. A conceptually simple procedure for comparing the performance of the four calibration methods in the presence of different levels of random noise in spectra data has been introduced here. This procedure, using the computer simulation data and real spectra of tobacco, has provided useful information for understanding the effects of random noise on the performance of multivariate calibration methods. Both numerical and graphical results will be shown.
© 1998 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