Linking Sensory and Consumer Data: Multivariate Modeling and Visualization Strategies for 'Short and Wide' Data

• Activity Number: 135
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Consulting
• Author(s): Jason Parcon*+
• Companies: PepsiCo
• Keywords: PLS Regression ; Multicollinearity ; Short data ; Sensory Research ; Consumer Research

Abstract

An important step in sensory and consumer research is to link sensory information on a set of test products and consumer assessments on the same set of items. This process allows the identification of sensory attributes that impact consumer value perception. The exercise is not trivial because of two main constraints: (1) available data are typically "short and wide", i.e., the sensory attributes far outnumber the products under investigation, and (2) the high level of multicollinearity among sensory attributes. This presentation will discuss data modeling strategies with the aforesaid constraints. The strategies typically involve dimension reduction followed by multivariate regression of consumer hedonic assessments versus sensory information. The said techniques have proved useful on several fronts: (1) they provide models that easily pinpoint the sensory markers that impact consumer liking, and (2) they provide tools for easily visualizing the relationships among multivariate information. The discussion will be aided by real (coded) data sets.