Correspondence analysis for check-all-that-apply data
In many sensory consumer studies, assessorscharacterize product samples by checking terms from a predefined list. Thesecheck-all-that-apply (CATA) responses are often coded so 0 and 1 indicate notchecked and checked respectively. The data are then organized into abinary array with three modes: assessors (blocks), products (objects), andterms (variables). Existing correspondence analysis procedures explorerelationships between objects and variables in CATA data by operating on aproduct-by-term frequency table obtained by collapsing the CATA array overassessors. This manuscript proposes a new correspondence analysis procedurecalled cataCA that operates on the CATA array, not on the frequency table. Theproposed cataCA procedure can be used to investigate CATA data from studieswhere each assessor evaluates each product once. The cataCA procedure isdescribed, then applied to consumer data from a CATA study. Various ways to visualize results from the cataCA procedure are illustrated. Analysis shows the relationship between three procedures: the cataCA procedure, the multiple-response correspondence analysis procedure, and the procedure of principal component analysis of a product-by-term table of citation rates with centred, unstandardized column variables. Limitations with the cataCA procedure are also highlighted. Recommendations are provided for which solution to prefer indifferent scenarios. R code for conducting the proposed cataCA procedure is provided.
Castura,J.C. (2026). Correspondence analysis for check-all-that-apply data (Version v1). Zenodo. https://doi.org/10.5281/zenodo.23322975