Modeling the relationship between sensory attributes and liking using generalized additive mixed models: a case study of plant-based meat alternatives.

This study compared two analytical frameworks, partial least squares regression (PLSR) and a generalized additive mixed model (GAMM), to analyze the relationship between sensory attributes and consumer liking in plant-based meat alternatives. Consumer liking was measured using a nine-point hedonic scale, and sensory intensities were obtained using the rate-all-that-apply method. Model performance was assessed using a nested cross-validation framework with multiple validation structures. Across all validation structures, the ordinal GAMM showed higher predictive performance than PLSR, accounting for ordinal ratings and nonlinear relationships. Both approaches identified consistent key sensory drivers of liking, including smoked flavor, meaty flavor, and beany odor. GAMM also revealed curvilinear relationships with optimal intensity ranges and interactions between attributes. Overall, PLSR efficiently summarizes attribute importance under near-linear conditions, whereas GAMM captures nonlinear dynamics and interactions. The complementary use of both approaches enables a more comprehensive understanding of sensory drivers of consumer acceptance.