Click on “Download PDF” for the PDF version or on the title for the HTML version.


If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

STUDY ON OUTLIER SAMPLE ELIMINATING CRITERIONS AND METHODS FOR BUILDING CALIBRATION MODEL OF NEAR INFRARED SPECTROSCOPY ANALYSIS

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  Paper number  036100,  2003 ASAE Annual Meeting . (doi: 10.13031/2013.14197) @2003
Authors:   Shiping Zhu, Yiming Wang, Xiaochao Zhang, Jingzhu Wu
Keywords:   NIR, PCR, PLS, Outlier Detection, Callback Arithmetic Operator

A new method is presented in this paper, which can be used to detect and eliminate outlier samples in building a PCR (Principle Component Regression) or PLS (Partial Least Squares) calibration model of NIR (Near Infrared Spectroscopy). Because of the existence of outlier samples in training sets, the predicted concentrations of samples are little accurate even erroneous, so it is very important to detect the outlier samples and eliminate them from training set. One usual criterion of outlier detecting is based on predicted concentration residuals, samples that have significantly larger concentration residuals than the rest of the training set are concentration outlier samples. Some non-outlier samples are often falsely regarded as outlier samples when using Outlier Detecting Once (ODO) method. Instead of ODO method, a new method using Callback Arithmetic Operator (CAO), named Outlier Detecting Twice (ODT), is developed. Using ODT method, more effective samples will be kept in final model than using ODO method; this made the final model to be more representative, stable and robust.

(Download PDF)    (Export to EndNotes)