Effective Calibration Transfer via M\'obius and Affine Transformations

Autor:	Kneale, Casey, Booksh, Karl S.
Rok vydání:	2017
Předmět:	Statistics - Applications Statistics - Methodology
Druh dokumentu:	Working Paper
Popis:	A novel technique for calibration transfer called the Modified Four Point Interpolant (MFPI) method is introduced for near infrared spectra. The method is founded on physical intuition and utilizes a series of quasiconformal maps in the frequency domain to transfer spectra from a slave instrument to a master instrument's approximated space. Comparisons between direct standardization (DS), piecewise direct standardization (PDS), and MFPI for two publicly available datasets are detailed herein. The results suggest that MFPI can outperform DS and PDS with respect to root mean squared errors of transfer and prediction. Combinations of MFPI with DS/PDS are also shown to reduce predictive errors after transfer. Comment: To be revised and submitted to the Journal of Chemometrics
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/1701.09143 Zobrazit plný text záznamu View this record from Arxiv