Main content area

Utilization of agglomerative hierarchical clustering in the analysis of hedonic scaled consumer acceptability data

Schilling, M.W., Coggins, P.C.
Journal of sensory studies 2007 v.22 no.4 pp. 477-491
analytical methods, cluster analysis, consumer acceptance, consumer preferences, fluid milk, food choices, ham, ingredients, research projects, shrimp
Agglomerative hierarchical clustering was utilized to group consumers together based on product preference and liking in four consumer-based sensory studies. This statistical technique was effective at determining variations in consumer preference as a result of both processing techniques and ingredient incorporation. Results revealed that agglomerative hierarchical clustering can often improve the interpretation of consumer sensory data when compared to currently utilized analyses, and has significant applications in research projects with a sensory component. Three recommendations for conducting a comprehensive statistical analysis of hedonic scaled consumer data are: (1) perform a randomized complete block design to test treatment effects using the total data set; (2) utilize agglomerative hierarchical clustering to group panelists based on preference and liking of food products; and (3) perform randomized complete block designs within each cluster. If significant differences occur among treatments within a cluster, use a mean separation technique to determine significant differences among treatments within that cluster. Agglomerative hierarchical clustering was coupled with traditionally used analyses in the evaluation of hedonic scaled consumer data pertaining to chicken nuggets, retorted ham, fluid milk and cooked shrimp. Coupling of cluster analysis and traditional analyses was effective at grouping consumers together based on product preference and liking. Randomized complete block designs were also utilized within each cluster for further differentiation among treatments. A full description on how to analyze hedonic scaled sensory data using agglomerative hierarchical clustering, randomized complete block designs and Fisher's least significant difference test was included in this research paper and is an effective analytical method for the evaluation of hedonic scaled consumer data.