The Northern Ireland group held a meeting at 4pm on Thursday the 23rd of May, 2013 in the David Bates Building in the Queen's University of Belfast. The speaker was James Nicholson of the Smart Census Centre, Durham University, UK.
The talk dealt with the use of interactive tables rather than static tables as an aid to interpreting complicated Census data. James motivated the main line by showing some relatively complicated, typically, 3-dimensional sets of data and discussing how their display might be improved upon.
He felt that complicated multivariate data were off-putting, especially since many people were not confident in mathematics and/or statistical modelling. Moreover, Census data were likely to be more complicated than data met routinely in practice. So the question was how to make quantitative data more accessible to people: students, the general public, politicians and policy-makers? In particular, could we create tools which enabled people to explore the data structure themselves and in a relatively short time?
He then illustrated software which had been developed at the Smart Centre in Durham University (www.smartcensus.org.uk) which aimed to achieve these goals. He emphasised that the software was still in development and asked for feedback.
The software allowed the user various options to manipulate the information displayed flexibly using sliders without loss of information. As demonstrated, it was limited to 3-dimensional data comprising one continuous Y variable (number or percentage response) and two categorical variables, say X1 and X2. One example was the percentage of people having a long-term illness broken down by age and religion. Some questions emerged at this juncture particularly in relation to regression modelling but this was not yet incorporated.
James reminded the audience that typically, the census numbers were very large, whence most effects would be statistically significant so that visual interpretation in terms of the magnitudes of differences would not systematically mislead. In fact, it was hoped that the story told by the data could be revealed quickly in this way without the need for more formal analysis.
The last part of the talk dealt with possible future applications for this type of software in tables with different structures and maps of health data. Meanwhile we were encouraged to feed back our ideas to the SmartCensus group.
His talk attracted great interest from members of NISRA who had attended the meeting in strength and a detailed discussion ensued.