Journal article
Calibration diagnostic and updating strategy based on quantitative modeling of near-infrared spectral residuals
Analyst (London), Vol.140(3), pp.786-796
02/07/2015
DOI: 10.1039/c4an01849d
PMID: 25473807
Abstract
A diagnostic and updating strategy is explored for multivariate calibrations based on near-infrared spectroscopy. For use with calibration models derived from spectral fitting or decomposition techniques, the proposed method constructs models that relate the residual concentrations remaining after a prediction to the residual spectra remaining after the information associated with the calibration model has been extracted. This residual modeling approach is evaluated for use with partial least-squares (PLS) models for predicting physiological levels of glucose in a simulated biological matrix. Residual models are constructed with both PLS and a hybrid technique based on the use of PLS scores as inputs to support vector regression. Calibration and residual models are built with both absorbance and single-beam data collected over 416 days. Effective models for the spectral residuals are built with both types of data and demonstrate the ability to diagnose and correct deviations in performance of the calibration model with time.
Details
- Title: Subtitle
- Calibration diagnostic and updating strategy based on quantitative modeling of near-infrared spectral residuals
- Creators
- Hua Yu - Department of Chemistry and Optical Science & Technology Center, University of Iowa, Iowa City, IA 52242, USA. gary-small@uiowa.eduGary W Small
- Resource Type
- Journal article
- Publication Details
- Analyst (London), Vol.140(3), pp.786-796
- DOI
- 10.1039/c4an01849d
- PMID
- 25473807
- ISSN
- 0003-2654
- eISSN
- 1364-5528
- Language
- English
- Date published
- 02/07/2015
- Academic Unit
- Chemistry
- Record Identifier
- 9984083806602771
Metrics
12 Record Views