INTRODUCTION
In the majority of sensory problems we are interested in
the average scores given by expert panels or by trained
panels. The variation between assessors should be minimized
during the selection of the members of the panel or
in the training period.
In the case of consumer panels, the reality is usually
quite different. For many reasons, I think that such
panels should rather be used for ranking or other types of
non-parametric methods.
When reading the vast recent literature covering this
area however, the use of numerical scores coupled with
analysis of variance appears to be very popular for both
types of panels.
In many research papers a description of the form of
the analysis of variance is often ignored or only very
briefly mentioned. Terms like ‘we used a factorial
design’, or perhaps more often ‘an analysis of variance
was used’ are rather common. It is difficult for the reader
to find out how the analysis was done. To reach a reliable
conclusion from the experimental design and the