Data visualization in Postscript
I produce the graphs that visualize quantitative information in my
papers by writing encapsulated Postscript (EPS) files with an editor.
This is relatively easy once you have some templates to work from, so
here are some of the files I have produced. You can see them as
picture and as program text on this page.
- pic text: line graph, constant x stride, with
median-of-3.
- pic text:
line graph, with x coordinates, separate line tags, logarithmic X
scale.
- pic text:
line graph with different Y scale factors for different lines.
- pic text: bar
graph, automatic colours (gray only) and x coordinates, separate
colour tags.
- pic text: bar graph, automatic non-gray colours
and x coordinates, separate colour tags.
- pic text: bar graph, automatic pattern colours
and x coordinates, separate colour tags; everyone I showed this to
said he found it visually irritating, so I went with non-pattern
colours instead.
- pic text: cumulative bar graph, automatic non-gray colours.
- pic text:
boxplots with outliers.
If you use RGB colours, take a look at this RGB
colour chart on different output devices (they usually come out
differently).
Some tips
Always store everything needed to get the data somewhere; you may need
to reproduce it later, e.g., in order to trace why some newer
experiment gives unexpected results. Among other things, you need to
remember what exactly you were measuring to get the results, and what
the results are. Sometimes a good place to remember these things is
the actual .eps file; e.g., put the command lines that produced the
results in comments, and put the output there, too, commenting out
unneeded data or unselecting it programmatically (e.g., with
median-of-3).
Anton Ertl