Univariate and multivariate modelling of flavour release in chewing gum using time-intensity: a comparison of data analytical methods

Sensory properties of chewing gum were measured using time-intensity (TI) methodology to record responses from assessors. Studying the information retrieved from a total of 800 TI curves, old and new data analysis methods were compared. These included the normalization method developed by Liu and MacFie, and ad hoc curve parameter retrieval, Principal Components Analysis (PCA) and ANOVA Partial Least Squares Regression (APLSR) methods. Two alternative methods have been developed for handling the multiple sources of variation in the data, including a dual PCA method and PARAllel FACtor analysis 2 (PARAFAC2) multiway analysis. A comparison of the different methods is conducted and the advantages of the different methods highlighted. All methods provide valuable interpretations of the time-intensity data and lead to similar conclusions. The multiway analysis provides a comprehensive overview of the data, but more experience with different datasets is needed to provide definite advices as to which methods are preferable.