Statistical methods for analysing near-infrared spectra in food authenticity stu
Statistical methods for analysing near-infrared spectra in food authenticity studies
“Is a sample of a particular type or not?” is the key question in food authenticity studies. The answer is difficult because while one sample type (authentic) is known, the alternative sample types (adulterated, mislabelled, etc) may not be defined at the outset. The focus of this statistical project is to address this core issue. Novel statistical methodologies that directly analyse near-infrared spectroscopic data in food authenticity studies will be developed. This approach will avoid the data reduction step, e.g. principal component analysis, which underlies previous studies. Methods that have high performance, even with very little training data, have recently been developed and these will be extended to the proposed direct classification and clustering methods.