Data Relationships and Multivariate Applications

Abstract

Multivariate statistics have found great application in all areas of quantitative sensory science. In this chapter we will briefly describe the two major work horses in the field: principal component analysis (PCA) and canonical variate analysis (CVA). PCA should be used with mean data and CVA with raw data, namely data including replicate observations. We also discuss generalized Procrustes analysis (GPA) which is used with free-choice profiling data as well as in any situation where one may want to compare the data spaces associated with multiple data measurements on the same products. Lastly we discuss (as a preliminary to further in-depth discussion in Chapter 19) internal and external preference mapping. We conclude by stressing that multivariate analyses should always be performed in conjunction with univariate analyses.