Compusense Research

Sensory studies aim to understand how products (e.g. food and beverages) are perceived and conceptualized. Measuring products analytically provides information about their physical and chemical properties. It is useful for product developers to have a more complete understanding which physical and chemical properties are associated with which sensations. Component-based methods are a popular way of exploring relationships between sensory and instrumental variables. This talk will focus on some recent advances in supervised principal component regression (SPCR) and partial least squares regression (PLSR). In many cases, the researcher is interested in modeling all objects. As will be shown, SPCR of all objects is equivalent to the SPCR of all paired comparisons. PLSR of all objects is equivalent to the PLSR of all paired comparisons. In some cases, however, the researcher is interested in understanding test-control paired differences, i.e. differences between test products, such as product prototypes, and a control product, such as a benchmark or reference product. When only test-control paired differences are of interest, SPCR and PLSR can be used to investigate relationships between sensory and instrumental data. SPCR of test-control paired differences is not equivalent to the SPCR of all paired comparisons. PLSR of test-control paired comparisons is not equivalent to the PLSR of all paired comparisons. The approach used for model fitting when only test-control paired differences gives a deeper understanding of these component-based methods.

Castura, J.C. (2025). Exploring the relationship between sensory and instrumental data with component-based methods. Korean Society of Food Science and Technology International Symposium and Annual Meeting. 2-4 July. Gwangju, Korea. (Invited Oral).