Monte Carlo based test for inferring about the unidimensionality of a Brazilian coffee sensory panel

Abstract

Sensory science uses the human senses as instruments of measurement. Sensory analysis makes it possible to study organoleptic properties of products using a panel of assessors. Kermit and Lengard (2006) say that a good sensory panel should provide results that are accurate, discriminating and precise. Thus, in a successful analysis, it is key to have a set of robust tools for monitoring individual assessor’s performances as well as the performance of the panel as a whole.

The success of using a sensory panel depends on its performance, i.e., its ability to identify small differences between products in certain attributes with statistical significance. A good panel performance is achieved when each panelist discriminates between products (large product variability), repeats the assessments (small within-assessor variability) and agrees with all other panelists on the sensory sensation that is described by a particular attribute with certain strength (small between-assessor variability) (Derndorfer, Baierl, Nimmervoll, & Sinkovits, 2005).

Evaluating unidimensionality of panel is essential for the efficient use of the sensory analysis. For such purpose, the Monte Carlo based test revealed to be powerful for large panel and sample sizes. It can be adopted as a tool for checking the efficiency of the training process of a panel.