Outlier and group detection in sensory panels using hierarchical cluster analysis with the Procrustes distance

Comparison of acceptability of tropical fruit mixed nectars by mean test and multivariate statistical analysis

“…Acerola (Malpighia emarginata D.C.), on the other way, has low sensory appeal and short fruit shelf-life but high nutritional value, mainly vitamin C and carotenoids, antioxidant pigments whose combination is responsible for the fruit red color (SOUSA et al, 2010;LIMA et al, 2005). Due to its excellent attractive color and taste combined with good nutritive and functional characteristics, mango (Mangifera indica L.) has gained great economic importance, ranking among the ten most planted cultures in the world (DEMBITSKY et al, 2011). Mangoes contain high levels of bioactive compounds as phenolic compounds and carotenoids (POOVARODOM et al, 2010).…”

“…The process of grouping involves basically two steps. The first is related with an estimation of one measure of dissimilarity between the treatments, and the second with the adoption of a technique of grouping for the group formation (QANNARI et al, 1997;DAHL;NAES, 2004). These results, when visualized through diagrams known as dendograms indicate similarity between samples, such as that a minor distance between points indicates larger similarity between them.…”

Identifying outlying assessors in sensory profiling using fuzzy clustering and multi-block methodology

“…For instance the GCA which can also be formulated within the context (Van der Burg and Dijksterhuis (1996)) of distances, can easily be implemented by just using the same weighting technique that is used for the Tucker-1 method above The Procrustes analysis could also be incorporated within this framework by using the weighted version of GPA for computing the distances. Note that the latter is somewhat similar to the method proposed in Dahl and Ns (2004) for clustering of sensory assessors. In that paper the Procrustes distance combined with hierarchical clustering was used to identify outlying assessors.…”

“…If not detected and eliminated, or weighted down, outlying assessors may lead to loss of useful insight and poor prediction performance (see e.g. Dahl & Ns, 2004). A number of useful graphical methods have been developed for this purpose (see for instance Ns, 1990, Brockhoff & Skovgaard, 1994.…”

“…This type of methods may be very important as a starting point for identifying those assessors that need further attention and also for making more explicit judgements about outlyingness. Some recently published methods that have considered some of these issues are Dahl and Ns (2004), Wakeling, Raats, and MacFie (1992) and Qannari, Courcoux, and Vigneau (2001).…”