Figuring It Out

Roemer J. Janse

Figures are Fundamental

Allow us to swiftly convey a message
Easier to compare than numbers
Can capture your audience

Island	Species	Mean flipper length (mm)
Biscoe	Adelie	188.8
Biscoe	Gentoo	217.2
Dream	Adelie	189.7
Dream	Chinstrap	195.8
Torgersen	Adelie	191.2

Figures are Tricky

Many styles to choose from
Much freedom aesthetically
Not a major focus of research

Colour Coding

Colouring for Colourblindness

Not all colours are equally distinguishable
Colourblindness can quickly make figures unreadable
More than one type of colourblindness exists
Printing in black-and-white also makes you colourblind

Colouring for Colourblindness

Let’s compare some different colour scales:

Colouring for Colourblindness

Green-blind (deuteranopia):

Colouring for Colourblindness

Red-blind (protanopia):

Colouring for Colourblindness

Blue-blind (tritanopia):

Colouring for Colourblindness

Monochromacy (achromatopsia):

Colouring for Colourblindness

Some colour scales are more accessible than others (viridis, magma, IBM. Wong, Tol)
No colour scale is perfectly accessible (and red-yellow-green definitely isn’t)
Some journals/institutions require specific colours (corporate identity)
Printing in black and white is often not sufficient

We should always convey meaning with more than just colours!

Substituting for Shapes

Shapes with areas can use patterns:

Substituting for Shapes

We can still use colours with patterns:

Substituting for shapes

Or just go crazy…

Substituting for Shapes

Lines can use linetypes:

Substituting for Shapes

One-dimensional objects can shapeshift:

Contrasting Colours

Colour contrast is important!

We might want to show text in our figures
The contrast between text and background colour is important for readability

Contrasting Colours

Contrast ratio’s:

Species	All black	Mixed black-white	Labelled
Adelie	1.4	15.2	21
Chinstrap	3.5	6.0	21
Gentoo	8.2	8.2	21

Ideal is 7:1 or 4.5:1 for larger text

More Materials

Calculate contrast ratio’s: https://snook.ca/technical/colour_contrast/colour.html
Considering colours: https://blog.datawrapper.de/colors/
Grey as a go-to: https://visualisingdata.com/2015/01/make-grey-best-friend/
Considering colourblindness: https://blog.datawrapper.de/colorblindness-part2/#Colorblind-safe-color-palettes

Plotting Pitfalls

Tricky or Tinkering?

We have many design choices when making our figures
Some design choices may influence the message we convey
Other design choices influence whether we convey a message at all

3D < 2D

3D barcharts? Don’t.
3D piecharts? Don’t.
3D scatterplots? Perhaps…

3D barcharts and piecharts

Issues:

Distorted angles/bar heights
Difficult to gauge exact values
Extra dimension might be communicated through other means

3D scatterplots

Dual Axes

Dual axes can make it easy to mislead people (false correlations)
Especially a problem if the two axes are not monotonically related
In general they are harder to read

Instead:

Side-by-side graphs

Cutting Corners

Some y-axes have clearly defined ranges
- Probability
- Scale measures (e.g. SF-36)
Some y-axes have some sensible range (e.g. lab-values)
Some y-axes have a sensible limit at one side (e.g. height)
Cutting axes may make it seem like the effect is larger than it actually is