Data Visualisation: Theory
Aims for this section
By the end of this section, you should…
- Be able to recognise possible issues with scientific figures and be able to critically evaluate data visualisations
- Have a framework to build useful, objective figures to illustrate your results
- Have resources to further investigate different aspects of scientific figure making
Source material: Ten Simple Rules for Better Figures by Rougier, Droettboom and Bourne, 2014
Rule 1: Know Your Audience
Who will be reading it, in what context?
- You
- Your supervisor(s)
- General scientific or research audience
- Experts in your field
- Experts in a specific method you use
- The general public
- Policy makers
- Undergraduate students
- Scientific journal
- A general lecture on a research area
- An outreach event
- Updating funders
- Funding application
- Thesis examiners
What different requirements might these audiences/contexts have?
- Understand the needs and knowledge level of your audience
- Adjust the complexity of your visualisation to match
Consider:
- Figure type
- Annotations, labels
- Amount of data being shown
- Terminology and jargon
Rule 2: Identify Your Message
- The figure should express an idea quickly and succinctly
- Part of this is ensuring you know what the message is
- People can get stuck knowing they “need to make a figure for this chunk of results” but are not sure what they actually need to convey
- Come up with a one-sentence statement that captures your results, before building the plot
- Show your plots to your colleagues and see if they can deduce the key point
![]()
Trying to pick the right plot for your use-case
Browsing documentation galleries can be a good idea…
But be careful not to pick an obscure, strange plot just because it looks pretty!
Also, check and see what other researchers in your field are using…
Technical considerations:
- Font size
- Vector graphics vs. raster image
- Line weight
- DPI or resolution
- Colour profile: RGB or CMYK
Rule 4: Captions Are Not Optional
- Captions provide additional context for the figure
- They help the reader to interpret the figure correctly
- They should be treated like axes labels, a legend… e.g. essential!
What might a caption look like…
- In a published research article?
- On a conference poster?
- In a seminar or talk?
- On an outreach leaflet?
- In a policy document?
Rule 5: Do Not Trust the Defaults
- Default settings in your chosen plotting software may not suit your needs entirely
- Defaults may be outdated, not suited to your field of research, or not be accessible
- Adjust settings as needed - but also recognise when this might be a time-sink!
Of course, the solution might also be that a different plot shows your results more clearly… remember Rule 2: Identify Your Message!
Rule 6: Use Colour Effectively
- Colour can be an important asset in your scientific visualisations
- Must be used mindfully to avoid confusion
According to Edward Tufte 1983, colour can be either your greatest ally or your worst enemy if not used properly (Rougier et al., 2014).
Uses for Colour
- Adding an additional dimension to your figures (caution!)
- Adding clarity to different marker or line styles
- For emphasis
- For heatmaps
Low resolution image not intended to be scaled up…
![]()
Estimate subsidence of volcanic edifices on planets - using both colour and symbol shape to differentiate.
![]()
Topographical map of Mars; co-ordinates system not plotted to avoid discussion on Martian projection systems!
Pitfalls of Colour-Use
- Relying purely on colour to distinguish can cause issues with colour vision deficiencies, printing in greyscale, different screen settings
- Check if this will be an issue using an online tool such as Viz Palette
- Colour maps may not accurately represent your data - non-perceptually uniform maps
![]()
Rainbow colour maps such as “jet” have previously been popular despite their issues. “Plasma” is “perceptually uniform” but can obscure details in extreme values.
![]()
Cividis is both “perceptually uniform” and also appears the same to audiences with and without colour vision deficiencies, but has the same issues as “plasma” with obscuring features in plots with a wide dynamic range.
![]()
Spring is a terrible colour map, even if it’s pretty.
![]()
Sometimes you do need a plot that will pick up these sorts of features in the dark and bright areas!
![]()
It can be easier to see the issues in an image than in a map
![]()
Other packages such as “Cmocean” also provide well-designed scientific colour maps.
![]()
Other packages such as “Cmocean” also provide well-designed scientific colour maps… and some strange ones
![]()
Plot and analyse x-ray element maps
How can colour infer order?
Data can be encoded by other means aside from colour… which of these could be ordered or infer order?
- Size
- Shape
- Grouping
- Area
- Position
- Saturation
- Line pattern
- Line weight
- Angle
- Connections
Best Practice
It is a good practice to, where possible, avoid conveying information purely through color. You should always consider adding other ways to convey the same information besides just color.
Check that your colour choices will be easily distinguishable from one another
Ensure that colour does not infer order where there is none!
Use perceptually uniform colour maps whenever possible
Rule 7: Do Not Mislead the Reader
- Figures should represent data as objectively as possible
- Avoid creating misleading figures
This seems straightforward:
to make a good scientific visualisations, don’t do academic fraud
but actually, it’s easier to accidentally misrepresent data than you might think.
Unintentional misleading of your audience often arises through not fully understanding how you have visually encoded your data.
- Size
- Shape
- Colour
- Grouping
- Area
- Position
- Saturation
- Line pattern
- Line weight
- Angle
- Connections
![]()
Stacked barplots; both panels show the same data
![]()
Bubble plot, where “z” value is shown by both colour map and size
![]()
All of these represent the same data…
![]()
Both represent the same data…
![]()
Both represent the same data…
![]()
Both represent the same data…
Rule 8: Avoid “Chartjunk”
- Chartjunk refers to all the unnecessary or confusing visual elements found in a figure
- Avoid overloading a figure with too much detail
But do:
- Add enough annotations that the reader understands the figure
- Understand that this is subjective!
If your figure is confusing, try fixing it by removing something rather than adding extra labels!
![]()
Yikes.
- When should we use gridlines?
- When are annotations helpful?
- What about legends? Are they chart junk?
Rule 9: Message Outshines Beauty
- The key requirement of a scientific graphic is to communicate scientific results
- Aesthetics should not distract from the message
![]()
Jet is rubish, and pretty ugly… but Spring is also a terrible colour map, even if it’s pretty.
- I really dislike the look and colour scheme of this, but I have to admit it does a better job of representing my data
Tools and libraries within Python; * denotes library used in this course
- Usually Matplotlib is the best option if you want total control over every aspect of your plot, and want to build something that doesn’t look like any examples from library galleries
- You can use Matplotlib or Seaborn to create some simple, schematic-type plots
- This is one of the areas where a graphics program will probably outshine other tools!
