next up previous
Next: MEDIATION Up: ORGANISING INFORMATION GRAPHICALLY Previous: Plot Layout

Grid Layout

Grids are a compact way of arranging multiple displays of data, and the layout allows for easy comparison of displays across rows and columns. A grid-layout is a view that positions its subviews in a grid format. No restrictions are placed on the types of these subviews.

As with plots, grid layouts also support overlays where multiple subviews may be superimposed on a tile. This is useful, for instance, when showing point clouds with superimposed fits. Grid layouts, then, offer a very general way of arranging arbitrary subviews in a grid format. In many common applications, there is a consistent pattern across the subviews, where all are of the same type but show different variate or case subsets from a dataset. We have designed some particular kinds of grid layouts for these situations, as summarised in the table below:

Layout type Data Format Default subview
1d-layout n variates row/column of n boxplot-view
xy-layout n x-variates, m y-variates m rows, n columns 2d-pointcloud
pairs-layout n variates n rows, n columns 2d-pointcloud
      label on diagonal
batch-layout data subsets k=1 row/column boxplot-view
  eg levels of k factors k > 1 grid  
table-layout data subsets One tile per subset bar
    Tiles are scaleable  
    Default: size $ \propto$ subset size  

Like plots, grid-layouts need a constraint system or ``glue'' to ensure that meaningful comparisons between the subviews are possible. The constraint system used by grid-layout is necessarily richer than that of plot; it has options which require all, or none, of the subviews to have the same horizontal (or vertical) extents, furthermore constraints can be imposed separately on views across rows or down columns.

The grid-plot is a special kind of plot, where the interior view is required to be a grid-layout. The grid-plot acts as a wrapper around grid-layouts that will provide associated axes and labels, in the same way that the basic plot does for the 1 and 2-d views; in fact a scatterplot matrix, is actually a grid-plot with a pairs-layout as its interior view. Grid plots have a constraint system similar to grid-layout, except here the constraints apply to the margin views of the grid plot and the subviews of the grid layout.


next up previous
Next: MEDIATION Up: ORGANISING INFORMATION GRAPHICALLY Previous: Plot Layout

2000-01-28