ABSTRACT
The sensory properties (SP) of foods are usually measured by: (a) naive consumer panels who express their acceptability for the food samples, and (b) trained panels who measure sensory properties analytically. In both cases a number of people have to be motivated, assembled, and instructed, and after the evaluations are completed, a large array of data is left behind to be statistically analyzed and interpreted. Physico-chemical measurements (PCM) are carried out by laboratory instruments which are a lot simpler to handle than people because they provide precise data with ease and speed. However, their drawback is that they may not predict sensory properties since these are ultimately perceived by the consumer, and thus this type of evaluation very often becomes useless. Regression tools to analyze the correlation between PCM and SP go from very simple linear equations of the form:
Sensory = a + b • Concentration
to linearized versions of psychophysical laws such as Steven’s law (Moskowitz, 1983):
Log(Sensory) = k + n • Log(Concentration) to multivariate data where a matrix of SP is correlated to a matrix of PCM. Partial Least Square Analysis (PLS) is a method especially developed for the correlation of multivariate data.
