This paper presents the application of correspondence analysis (CA) for investigating associations using confidence regions (CRs) with a focus on facilitating mining the data and hypothesis generation. We study the relationship between locations and "less-healthy" food consumption by UK teenagers. CA allows for a quick visual inspection of the various association structures that exist between the categories of cross-classified variables in large datasets derived with varying study designs. The hypotheses generated by the visual display can then be independently tested using suitable regression models. CA makes use of readily available software tools and of robust statistical tests amenable to interpretation.
Keywords: Confidence Regions; Correspondence Analysis; Hypothesis generation; Hypothesis testing; clustered data; food-group; healthy; hierarchic data; location.