Analysis of chemical compounds in beverages – Guidance for establishing a compositional analysis. (30th September 2020)

• Record Type: Journal Article
• Authors: Templ, Matthias; Templ, Barbara
• Is Part Of: Food chemistry. Volume 325(2020)
• Publication Date: 2020-09-30
• Subjects: Beer -- Forced aging -- Chemometrics -- Log-ratio analysis -- Correlation -- PCA -- Misclassification; Food -- Analysis -- Periodicals; Food -- Composition -- Periodicals; 664
• DOI: 10.1016/j.foodchem.2020.126755
• Languages: English
• ISSNs: 0308-8146

Abstract

Highlights: Correlation analysis conducted on raw data may lead to spoiled results. Classical PCA of beer data resulted in distorted biplots and less explained variance. Classification resulted in less predictive power when non-CoDa method was used. Compositional Data analysis (CoDa) of chemical compounds enables the correct assessment of compositions. Aged beer variation is driven by Furfural concentration. Abstract: In recent years, much analysis has been carried out regarding the chemical compositions of nonalcoholic and alcoholic beverages. However, no study has ever considered the compositional pitfalls inherent in such studies, which may lead to arbitrary results. Two approaches were compared in this paper, compositional data analysis (CoDa) and classical statistical analyses, to demonstrate (1) how aging affects beer and (2) how the results vary depending on the applied approach. Exemplarily, chemical compounds of 43 beer samples were analyzed. Our work has led us to obtain different results when the correlation analysis and principal component analysis were conducted on the unmodified-, log-transformed-, or closed-data in contrary to the analyses of log-ratio coordinates. The outcome of the latter provided better separation between aged and fresh beer samples and also easier interpretability. These results lead us to advise the usage of CoDa methods for the analyses of chemical compositions of beverages.