Improving the display of correspondence analysis using moon plots

Tim Bock

Numbers International Pty Ltd


Correspondence analysis is available in most general-purpose statistical software. It is probably the most widely used multivariate technique for analysing tabular data. It is often used by practitioners conducting brand positioning and segmentation studies, with its output showing a ‘map’ of a market.

Despite its popularity, correspondence analysis plots are easily misinterpreted. This paper presents a new plotting method that reduces some of the potential for misinterpretation. The paper is structured as follows: first is a description of the mechanics of correspondence analysis; second is a review of the main possibilities for misinterpretation; and third is the introduction of the new plot.