A new approach to analysing TDS data using GPA, CLUSTATIS, and AHC

Generally, interpretation of Temporal Dominance of Sensations (TDS) data is based on the visual inspection of dominance curves, but there is a growing recognition for the need to identify additional approaches. This study investigated the application of General Procrustes Analysis (GPA), CLUSTATIS, and Agglomerative Hierarchical Clustering (AHC) for the analysis of TDS data, using bacon ( n = 8) and cooked ham (n = 8) products as case studies. Using the data proportions obtained by the TDS, the first approach (i.e. GPA) focussed on exploring the global sensory dynamics of each product category throughout oral processing. In the second approach, dendograms were constructed using CLUSTATIS and compared to those obtained by AHC to determine the extent to which CLUSTATIS considered the temporarily of the dataset and to explore which of the two techniques is more effective at clustering similar products. GPA revealed the sequence at which the different attributes peaked during oral processing within each product category. Notably, during the initial tasting phase of both products, perception of texture attributes peaked, followed by the emergence of flavour attributes, while taste attributes peaked towards the end of the evaluation. Both AHC and CLUSTATIS enabled the clustering and comparison of more than two products simultaneously, which is not possible with conventional approaches. Although CLUSTATIS considered the temporality of the data, AHC proved more effective at clustering identical products across different repetitions. However, the possibility of low repeatability across repetitions should be considered. Overall, applying these approaches revealed new patterns in the temporal sensory characteristics of the products and in their perception. • GPA on TDS data visualized the tasting dynamics of food consumption. • Tasting dynamics: the sequence at which attributes peak during consumption. • CLUSTATIS considered the temporality of TDS data. • AHC clustered TDS data based on the intensities of the attributes.