Outlier Sample Eliminating Criterions and Methods for Building Calibration Model of Near Infrared Spectroscopy Analysis

祝诗平,王一鸣,张小超,吴静珠

DOI: https://doi.org/10.3969/j.issn.1000-1298.2004.04.031

2004-01-01

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