Scaling Up Graphics
The design and implementation of statistical graphics should pay attention to the challenges from big datasets. For many users, this has not been an issue up till now and so some statistical and graphics packages can have problems with graphics of more than 10,000 cases.
However, most of the plots used in statistical graphics can be scaled up to be usable with large datasets. Areal plots for categorical data are quite robust against large data glyph-based plots do have more serious problems. Modifications like α-blending or binning, interactions like (logical) zooming and panning, or interactive reordering and grouping are of great assistance when dealing with large datasets.
In general, all statistical graphics that summarize the data, and plot some version of these summaries, will scale up to large datasets. Barcharts, for instance, plot the breakdown of a categorical variable, which is a sufficient summary to fully describe the data. Binned scatterplots show an approximation of the underlying scatterplot and have a complexity that depends on the (constant) size of the binning grid rather than on the size of the dataset.
Unable to display preview. Download preview PDF.