Analysis and Interpretation of Sensory Descriptive Data In this paper I assume that during the training and testing phases of the experiment all appropriate sensory techniques had been performed properly. In other words, the panel was well trained, had access to the appropriate reference standards and had input into the number of samples evaluated per session. Additionally, I assume that the samples were served according to an appropriate experimental design, namely that all samples were evaluated in at least duplicate or, preferably, triplicate by all panelists in a randomized design. I will concentrate on the analysis and interpretation of the data from such a test. The data analysis will usually start with univariate analyses of variance (ANOVAs) of the terms (attributes) evaluated by the panel. Since we had replicate judgments1we should do a three-way analysis of variance with main effects: panelist, wine, replication; and interaction effects: wine*panelist, wine*replication and panelist*replication. Table 1 is an example ANOVA table. In this case a fixed effect analysis was performed (all main effects were fixed and we cannot extrapolate our results beyond this panel and these wines). Table 1: Summary of the fixed effect three-way analysis of variance of floral aroma intensity ratings for wines fermented with four different strains of Saccharomyces cerevisiae.Source of Variation df MS F-value Significance* Panelists (P) 9 6.15 24.10 * Wine (W) 3 50.81 199.21 * Replications(R) 1 0.01 0.05 ns P*W 27 7.78 30.49 * P*R 9 0.43 1.68 ns W*R 3 0.25 0.46 ns Error 27 0.26 _____________________________________________________________________ ns, and * denote, respectively, no significant difference and differences significant at p<0.05 For the main effects, the wines were significantly different in perceived floral intensity and the panelists were a significant source of variation – without looking at the panelist interaction effects we do not know whether the panelists’ variation was due to the normal individual differences or also due to panelists not being consistent with themselves and with each other. The replications were not different – this is good since one would assume that the bottles opened for the various replicates should not differ. The interaction effects tell us that a) since P*R is not significant the panelists were consistent from replication to replication; b) since the W*R is not significant the wines did not differ 1If there were no replications, a two-way ANOVA with main effects panelist and wine should be performed. A two-way ANOVA with no interaction effects cannot tell us if the judges were consistent or reproducible.
has intentionally blurred sections.
Sign up to view the full version.