MODELING DIFFERENCES BETWEEN PANELISTS IN USE OF MEASUREMENT SCALES

In sensory data sets, an important source of differences between panel ists is in their use of the measurement scale. These differences can b e summarized in differences in location, the overall level, and differ ences in dispersion, the range of the scale used. This paper discusses a method of correcting for these differences by jointly modeling loca tion and dispersion using a see-saw algorithm. This approach is also a pplicable when scores are not normally distributed and when there is a (nonlinear) relationship between the dispersion and the location. The approach is illustrated with an example for flavor data of freeze-dri ed and hot-air dried peppers.