3-Way and 3-block PLS regressions in consumer preference analysis

PLS regression modelling using 3-way or 3-block data is sometimes misleadingly regarded as a single method. This paper illuminates the structural differences of 3-way data compared to 3-block data applied to sensory analysis. The data used to illustrate these two powerful methods is a set of 17 tomato varieties with information acquired from chemical measurements, sensory panel evaluations and consumer likings, attitudes and demographics. A 3-way PLS regression model is used to identify positive and negative drivers of consumer likings and simultaneously evaluate the performance of individual sensory panelists. From the 3-block PLS regression model, also called L-PLS model, the tomato varieties are described using chemical information and sensory attributes. Also extracted from the L-PLS model is background information on the consumers who expressed their preferences for the different tomato types.