January Webinar

As quantitative scientists we forget just how hard most people find dealing with maths and data.  This creates a huge problem if lay decision-makers can't understand our complex analyses.  At best, our analysis isn't used in the decision-making process, at worst it gets corrupted and abused.

Data Visualisation is one solution to the problem, where we can leverage the massive power of our visual processing system to translate complex analysis into simple, intuitive insights.  Large strides in producing user-friendly graphs and charts have been made by prominent visualisation experts such as Edward Tufte, William Cleveland and Stephen Few.  But delving into the data viz literature, much of what we "know" about data visualisation is based on practitioner experience and opinion - there's surprisingly little quantitative research into the subject.  What there is has focused almost exclusively on accuracy, but this is just one factor that makes a chart or graph successful (others include ease of use, the correct choice of physical analogue and good design).

For example, in the field of Data Analytics, the hatred for pie charts is near ubiquitous.  However, this seems to be based on the misconception that when looking at a pie chart we compare areas and angles, and that this is less accurate than comparing lengths (as in a bar chart).  But there is no consistent evidence in the literature to support this.

Of major importance is that there seems to be little research into the visualisation of probability distributions, a key output of many OR analyses. Andy will argue there is a need to rethink how we do research in the data visualisation field, so that we can design better graphs and charts for decision-making.