Multivariate Modeling of Aging in Bottled Lager Beer by Principal Component Analysis and Multiple Regression Methods

Abstract

Data collected from the sensory test score evaluation of bottled lager beer, together with the chemical components related to aging, including carbonyl compounds, higher alcohols, unsaturated fatty acid, organic acids, alpha-amino acids, dissolved oxygen, and staling evaluation indices, including lag time of electron spin resonance (ESR) curve, 1,1'-diphenyl-2-picrylhydrazyl (DPPH) scavenged amounts, and thiobarbituric acid (TBA) values, were used to predict the extent of aging in bottled lager beer, using both multiple linear regression and principal component analysis methods. Carbonyl compounds, higher alcohols, and TBA value were significantly and positively correlated with sensory evaluation of staling flavor. While lag time and DPPH scavenging amount were negatively correlated with taste test score. Multiple regression analysis was used to fit the sensory test data using the above chemical compound aging related parameters and evaluation indices as predictors. A variable selection method based on high loadings of varimax rotated principal components was used to obtain subsets of the predominant predictor variables to be included in the regression model of beer aging, so as to eliminate the multicollinearity of the original nine variables. It was found that staling extent was influenced significantly by higher alcohols, TBA value, and DPPH scavenging amount, and the multicollinearity of the regression model was found to be weak by examining the variance inflation factors of the new predictor variables. A mathematic model of the organoleptic test score for beer aging using these three predictors was obtained by multiple linear regression, showing that the major contributors to the sensory taste of beer aging were higher alcohols, TBA index, and DPPH scavenging amount, with the adjusted R(2) of the model being 0.62.