A peek into the world of data visualisation with Prof. Munzner

Visualization Analysis & Design – A tutorial

As you probably already know, I work in user experience in a complex data environment, so I feel that it’s important to know (at least superficially) about “Data Visualisation” (aka “vis”).  To this end, I recently attended a full day tutorial given by Prof. Tamara Munzner from the Department of Computer Science at the University of British Columbia, Canada.  I sketchnoted frantically, as there was a lot of ground to cover, and I used Tamara’s comprehensive slide sets (on her website) to finish the notes afterwards.

Read on if you want to quickly gen up on data vis

Below is a brief tour of what was covered in Tamara’s comprehensive tutorial.  There are five sketchnotes with a few explanatory notes under each.  The last one includes a case study of how a genetics vis tool was designed, using a light-weight UX-style process.  Enjoy!

(1/5) What is vis and what are the constraints on it?

munzner_vis_1_of_5

Definition of Visualisation

Visualisations are computer systems that provide representations of datasets designed to help people carry out tasks effectively.  80% of the time they are needed to make a data analysis task faster, and 20% of the time they solve new problems.

There are three kinds of constraints on visualisation: human, technical and mixed

  1. Human: the nature of the visual system, and our attention span!
  2. Technical: limits on display pixel-count (and sufficient corresponding white space to show things clearly), and computer processing power.
  3. Mix of human and technical: there are only so many practicable design idioms to choose from (see sketchnote 4 of 5 for more on idioms)

(2/5) What are the fundamentals you need to know before you start?

munzner_vis_2_of_5To design a new vis system there are three key steps

You need to find out:

  1. What is the data you are dealing with?  e.g. does it split up into categories?  Is it quantitative?  Is it ordered?, etc.
  2. (UX stuff:) Why does the user need this data?  What is the task they are trying to complete with it?
  3. How will your vis look? i.e which visual encoding (aka ‘idiom’) will work here?

Tamara showed findings from a paper that indicate perception is not equally reliable across different visual representations,  e.g. length is perceived accurately, but area, brightness and colour not as accurately.  Take home message: length may be the best option, where appropriate, and colours are to be used carefully!  Tamara’s handy crib sheet ranks them in order of effectiveness:

crib sheet

(3/5) Vis rules of thumb including colour and angles

munzner_vis_3_of_5Colour with caution!

When thinking about colour in vis, there are three things to consider:

  1. Hue: what we think of as colour, the type/category e.g. red, brown, purple.  You can only distinguish 10-12 max. in one vis.
  2. Saturation: like the dilution of the colour, how much of it is there?
  3. Luminance (also called brightness): describes the amount of light that is emitted from a particular area.  This varies with hue.

Size affects salience of colour

For high saturation colours use small areas, for low saturation colours use large areas.  For vis, RGB colour encoding is a bad idea, because it’s not the best choice for human perception.  HSL is a much better option, but be aware in this context lightness is not equal to luminance.

(4/5) Idiom design choices – how will you encode data visually?

Idioms

Idioms

Definition of an ‘idiom’ in data vis

An idiom is a distinct approach for creating or manipulating a visual representation, i.e. it includes:

  1. how to draw it: the visual encoding, for which there are many possibilities as shown in the above sketchnote (this is not exhaustive by the way!)
  2. how to manipulate it: interactions, such as animations pop-outs, change in colour hover-overs, scale/size adjustments, etc.

(5/5) Thinking about the users/usage of vis

munzner_vis_5_of_5

Paper about design study methodology

Reflections from the trenches and from the stacks” is the subheading of this paper – as it includes the real-life lessons learned by Tamara’s team after completing 21 separate vis projects.  The authors present a neat nine step framework for conducting a design study for vis systems.  There is also a natty little matrix to help you decide on the suitability of this proposed framework for your own projects.

User Experience meets vis: a case study

As far as I can tell, UX and Data Vis seem to operate as separate communities but judging by Tamara’s examples, cross-pollination is happening; for example, her group’s “Variant View” visualisation tool was developed using a light-weight user-centred design approach.  This tool is designed for analysis of gene sequence variants on a horizontal gene browser, like EMBL-EBI’s Ensembl or UCSC’s Genome Browser.   (This example really piqued my interest because it is exactly the sort of tool I work on at EMBL-EBI.)

Light-weight (‘agile’) UX design

Tamara’s approach involved carrying out semi-structured interviews with users every week to iteratively design and refine the vis.  The outcome looked simple and elegant, much like the websites people might use in their everyday lives – unlike the usual overly-complex web applications you often see in science.

Thank you Prof. Munzner et al.!

Well reader, consider your eyes now opened to the exciting world of data vis!  Thanks goes to Prof. Tamara Munzner for her succinct day course, which lifted the lid on an otherwise scarily complex subject.  The information on her website is so well organised, it is an inspiration and joy to visit.  Go there and see for yourself.  Juicy vis stuff is waiting for you there!

Useful Links

Tamara Munzner’s webpage (includes her bio, talk notes, etc.)

  • Slides on fundamentals: PDF 1up
  • Slides on idioms 1: PDF 1up,  2: PDF 1up
  • Slides on guidelines and examples: PDF 1up

Paper on Design Study Methodology: Sedlmair M., Meyer M. and Munzner T. (2012) Design Study Methodology: Reflections from the Trenches and the Stacks. IEEE Trans. Visualization and Computer Graphics (Proc. InfoVis), 18(12): 2431-2440 [PDF]

Paper on Variant View: Ferstay J.A., Nielsen C.B. and Munzner T. (2013) Variant View: Visualizing Sequence Variants in their Gene Context. Proceedings of IEEE Conference on Information Visualization, Atlanta, GA, USA.

Conferences for Biological Data Vis Community:

Nils Gehlenborg’s webpage (former EMBL-EBI-er now specialising in Biodata Vis research)

Data Viz sketchnote featured on the Sketchnote Army blog (March 2015)

 

Leave a Reply