Performance of Snow Flakes, Suns, and Factorial Suns in the Graphical Representation of Multivariate Data
Ingwer Borg, Thomas Staufenbiel · Multivariate Behavioral Research · 1992
The representation of multivariate data by icons has recently become more popular. The icon used most frequently is the snow flake which is available in most computer packages. This icon, however, has a number of drawbacks that are likely to lead to wrong interpretations. Similarity judgments on snow flakes can be dominated, in particular, by their size, a feature that varies depending on the arbitrary assignment of variables to rays. Another icon, the sun, reduces this problem by simply leaving the rays unconnected, but it too assigns the measurement variables to rays in an arbitrary way. A different type of icon, the factorial sun, is suggested here. It not only represents the values of the different variables, but also their correlational structure as much as possible through a 2-dimensional principal component plot of the variables' intercorrelations. Two experiments show that factorial suns do indeed lead to better interpretations by naive subjects than either snow flakes or (traditional) suns.