On the Edge: Statistics & Computing
What else every selfrespecting scatterplot should have?
Abstract
In a recent column in this series Duncan Murdoch (2000) described the benefits of writing your own scatterplot code. He mentioned the importance of specifying good scales; the flexibility ofhavingyourown software; the advantages ofoverlaying lines (or curves); the need for multiple plots; and, in an aside, the possibilities of having interactive tools for scaling and reformatting. Although that is a longish and a good list, it is in some ways more striking for what it leaves out than for what it includes, and this column discusses what else every selfrespecting scatterplot should have. Predominantly interactive features are suggested, something that is sadly missing from most statistical software. Many of the additions described would be advantageous in static displays for presentation graphics too, but it is much more effective if they are implemented interactively. Writing fully interactive software is challenging, but rewarding. The interactive tools that software should offer in general are discussed in Unwin (1999). Principles of GUI (Graphical User Interface) design relating to statistical software are outlined in Unwin et al. (1996). John Chambers has pointed