Panel performance for Temporal Dominance of Sensations

• Published in: Food quality and preference 2014-12, Vol.38, p.24-29
• Main Authors: Lepage, Melissa, Neville, Tara, Rytz, Andreas, Schlich, Pascal, Martin, Nathalie, Pineau, Nicolas
• Format: Article
• Language: English
• Subjects: Food and Nutrition; Life Sciences; Panel agreement; Panel discrimination; Panel performance; Subject agreement; Subject discrimination; Temporal Dominance of Sensations (TDS)
• ISSN: 0950-3293; 1873-6343
• DOI: 10.1016/j.foodqual.2014.05.002

•A methodology for assessing panel performance for TDS framework is proposed.•It is composed of two main elements: testing protocol+sound statistical analyses.•Performance results are summarized in a single table to have an easy-to-read output.•This approach does not require any pre-visualization step before execution.•The data analysis is based on ANOVA models logic+permutation tests. Given the particular nature of TDS data, specific analyses were developed and published over the past years to characterize the evolution of the dominant sensations in a product, to compare products to each other and to map the entire product space. Statistical approaches were also proposed to assess panel discrimination ability and agreement between subjects. To continue further on this direction, a general framework is proposed to evaluate panel and subject performance in TDS context. A protocol for testing products to evaluate the performance is proposed (before the measurement phase). Seven indicators are then described and summarized in a single table to facilitate the interpretation of the results by the panel leader. Three of the indicators describe panel and subject behavior in terms of attribute selection. The four remaining indicators describe the discrimination ability and the agreement, both at panel and subject levels. These indicators are inspired from usual ANOVA calculations used for descriptive profiling, but tested according to a permutation approach to overcome issues due to the nature of TDS data (non-independency between recorded values).