In a time where BigData (whatever that means) becomes more and more prevalent, the ability to find meaning in data by visualising it is a great asset. Indeed, it is something that I continuously try to learn and improve, and I’ll tell you about some inspiring visualisations as well as some lessons I have learnt on the way.
Looking both for advice and inspiration, I recently finished Nathan Yau’s Data Points: Visualization that means something. I got it after his first book, Visualize this, which is a good introduction to visualisation but also requires some work, which I did not always fancy doing. Data Points, on the other hand, is more of a tour of the kind of data visualisation that can be done. I enjoyed it immensely! The most impressive examples Yau brings up are quite far from what I usually plot – they are basically data art, like this wind map of the US. Nevertheless, he also introduces some of the nuts and bolts for graphically displaying data: what are the kind of plots you can make with a time series? Which kinds of visual cues are available, and what kind of data do they lend themselves to?
I’ll mention one point that Yau highlighted which I haven’t given much thought before: how good are we (or the readers of your graphic) at perceiving graphical representations of data? It turns out, we’re best at discerning position, and pretty good at length; getting worse for angle, direction, area, volume, saturation, and worst at hue. So to make something as clear as possible, pick from the early items on the list whenever possible.
All in all, I found this book very useful to take a step back from my data and the plots I always make, to think about why I make these plots, how I could visualise my data differently, and better.
Another lesson, quite literally, came from the edx course on statistical thinking for data science. Their module on data visualisations nicely touched on many basics that anyone making graphs should remember: What is it that you want to get across? Who is your audience? More tipps here and here. And a successful example – interactively showing the frequency of baby names (if you can ignore the pink and light blue stuff on the side!).
Another take on visualisation, or rather, science illustration, comes from a friend of mine who works as a freelance science illustrator. She describes the process of going from the science to her way of visualizing it. And she also points out some very useful and widely applicable advice on data visualisation from circos.
That leads me to the final point: If you are a scientist, I really hope that you know this paper on going beyond bar graphs (or already follow the authors’ advice anyway). How many times have you sat in a seminar where the speaker presented data in a way that was neither pretty nor understandable? Yeah, let’s change that.