A Checklist For Good Graphical Practice
From: Gordon
& Finch (2015) “Statistician Heal Thyself: Have We Lost the Plot?”,
JCGS, 1210-1229,
These questions and suggestions are a useful set to apply to graphs
designed for presentation purposes. I don’t agree with all of them, but
they are a very good start for things to think about in graphical
practice.
How clear is your purpose in communication?
- What relationships or patterns can you identify in the graph?
- Are these the relationships or patterns you intended to
represent?
- Can the viewer identify the patterns you wish to illustrate?
- Are the important comparisons you wish to show salient?
Make clarity a high priority.
- Does the graph have a clear title?
- Are the axes labeled?
- Are the units of the variables measured defined?
- Are the units of observation clear?
- Is the graph large enough?
- Would ordering groups or variables plotted improve the graph?
- Are all the graph labels horizontal?
- Use points to plot estimates (e.g., means, proportions) rather than
bars.
Consider detection issues.
- Can all the data points be seen?
- Are patterns in the data clear?
- Are the fonts large enough?
- Would it help to use jittering, or another form of representing
multiple, identical values?
Would panels help?
- Are there grouping variables that can be used to panel the
graph?
- Do the grouping variables correspond to the variation of
interest?
- Would additional panels help?
Align quantities to be compared on a common scale.
- Has distortion of the data been avoided by using the same scales for
the same measurement?
- Are measurements made on the same scale plotted on the same
scale?
- Would transposition improve the graph?
Does the graph have grid lines?
- Light gray grid lines will help with accurate interpretation.
Are all the elements of the graph defined?
- What do points on the graph correspond to?
- Are estimates (e.g., means, proportions) plotted on the graph
clearly defined?
- Are bars around points on the graph clearly defined?
How much decoding work does the viewer have to do?
- Is it easy for someone unfamiliar with your data to interpret your
graph?
- Does the graph stand alone?
- Try it on a friend!