PANCA: Panel concordance analysis

This paper proposes panel concordance analysis (PANCA) as a tool for panel leaders to identify disconsensus between the panelists on the sensory attributes used. PANCA summarizes the sensory data ([products × panelists × replicates] × attributes) by a low-rank approximation which is penalized for disconsensus (disagreement) between the panelists. When all the panelists agree on the sensory attributes used, the disconsensus penalty will have a negligible effect. However, if the assumption of good consensus is not supported by the data, considerable residual errors will arise. Consequently, PANCA can be used to identify difficult sensory attributes or even poor/deviating panelists which requires further training or could call for an alternative data processing strategy. It is also demonstrated that PANCA can be used to apply a multivariate ANOVA decomposition like in ASCA (ANOVA simultaneous component analysis). Theory and applications are explained by means of a real-life example from industrial sensory practice.