Interpreting sensory data by combining principal component analysis and analysis of variance

This paper compares two different methods for combining PCA and ANOVA for sensory profiling data. One of the methods is based on first using PCA on raw data and then relating dominating principal components to the design variables. The other method is based on first estimating ANOVA effects and then using PCA to analyse the different effect matrices. The properties of the methods are discussed and they are compared on a data set based on sensory analysis of a candy product. Some new plots are also proposed for improved interpretation of results.