Statistical methods for Check-All-That-Apply Data

Abstract

Check-All-That-Apply (CATA) question profiling has been proved to be a simple and convenient consumer profiling method to help the sensory researchers to understand how consumers perceive the food products. CATA profiling is gaining popularity in sensory and consumer science research due to its simplicity. The statistical methodologies on analysing CATA data in sensory and consumer science have been developing quickly recently. The investigating on new statistical methodologies to analyse CATA data is highly demanding and prioritised. There are three aspects in this thesis. 1. The statistical methods that have been applied on CATA data by the sensory researchers were reviewed. Also, the methods that were appropriate for analysing CATA data but unaware by the sensory researchers were introduced. The applications of both types of methods were demonstrated on three case studies that were collected from three different scenarios. At the end of the case studies, the procedures of analysing CATA data were suggested. 2. A CATA ballot contains a number of sensory terms for consumers to choose, but not all the prepared terms are useful to obtain information. The representative CATA terms enable the sensory researchers to have better understanding of how consumers perceive the products so that they can position the new products with clearer image. The methodologies of refining representative CATA terms were investigated. A strategy to refine the representative CATA terms was suggested. 3. The experiments often have limited budget which could impact the sample size. The simulation study showed how the reducing sample size affected on the statistical test results and the similarity to the original data. The representative CATA terms were included for the discussion of recommended sample size. This thesis showed that different statistical methods can be used to provide the answers to the research questions. At the end, the advantages and disadvantages of the statistical methods were discussed. Finally, it was recommended that the applications of distance based multivariate statistical methods should be considered by the sensory researchers, not only for analysing CATA data, but also for other consumer profiling data.