Advice on the Design and Analysis of Sensory Evaluation Experiments Using a Case‐study with Cooked Pork

Certain sensory evaluation experiments (e.g. taste panels) lead to data on an ordinal scale. Judges can typically only assess about six samples at each session before their ability to discriminate declines considerably. Thus allowance must be made for effects due to judge, session and order of tasting as well as treatment. We exemplify analysis-of-variance techniques where the basic observation is either a single assessment by a judge, the arithmetic mean over judges or a trimmed mean over judges. We also consider a linear logistic model. This specifically accounts for the ordinal nature of the data but the analysis and assessment of the goodness of fit by using residuals are more difficult.