Statistical analysis of sensory profiling data. Graphs for presenting results (PCA and ANOVA)

A principal component analysis is performed to analyse the matrix of median values, with 16 varieties in rows, and all descriptors in columns. Only texture descriptors contribute to the definition of the main axes. Six varieties are identified that score high on mealiness and low on moisture, the 10 other varieties are ordered according to a mashable axis which involves texture descriptors. Next we exclude texture descriptors since the main characteristics are already found, and mealy varieties since they are not suitable for steamed potatoes. A mixed linear model (ANOVA) with random subject effects is then used for each descriptor. Flavour and taste differences are found among the 10 varieties involved. Sweet, chesnut, pastry, cereal, earthy, herbace, rawpotat, artichok and celery descriptors differ between varieties (p<5%).