Share this on social media:

I have written approvingly, in an article for the June/July 2007 issue of Scientific Computing World, of the interface used by a statistics program designed for school use. My suggestion was that science can only benefit from the distribution of computerised analytic activity amongst larger sections of the population, that the publishers of statistics software are already looking in that direction, and that the designers of heavyweight packages could do worse than learn from such an interface. 

Until such time as top end tools acquire front ends accessible to lay users, however, there are products designed specifically to package exploration friendly controls, plain natural language help on screen, and only those methods useful to the intended user. Viz!on (yes, I know - just pronounce it ‘Vision’), the visualisation viewer from Data Description Corporation, is one such and gives another pointer towards how some aspects of a standard future office suite data analysis component might look. 

Viz!on uses Excel as its worksheet, and installing itself on the Excel menu bar. When clicked, it offers to handle one, two or three variables (a variable need not be a complete column, it can be any arbitrary contiguous subsection within one); it asks what combination of quantitative and/or categorical types are involved, and asks for the variables themselves to be chosen.Once those choices are made, a matrix of appropriate initial descriptive graphics is immediately generated. 

At every stage thereafter the screen carries small prompts which, when clicked, open up into tools and/or explanations. Every new tool opens with an accompanying note on what it is for, and full (but not tedious or over long) help is always only a click away. 

The facilities available concern, as the product name suggests, intuitive visualisation of summary and relationship measures. Histograms, scattergrams, normality plots, dot plots, simple correlation measures (Kendall’s Tau for nonlinear associations), data cleaning, selective colouring and data brushing, are provided in ways that make them usable without specialist knowledge or training. It would be a mistake to judge by weight, here: appropriateness for purpose, not extent, is the criterion. Data Desk, another product from the same company, expands the repertoire should that be required. 

If there is anything to the thesis that widening access to analysis benefits science, a product such as this should allow non-scientists to do something of use to scientists. The clinical director of a large hospital gave me anonymised body metric data on a hundred past female patients, which I passed to seven young fashion design students – intelligent, familiar with the human body shape, but lacking any post-compulsory science or mathematics. After an hour familiarising themselves with Viz!on, they identified twelve features in the data which seemed to them significant. The clinical director who supplied the data confirmed that all twelve features were valid and three were of serious medical interest – one of those three having been missed by the hospital’s own medical staff. 

There are some points of criticism, though they are minor and cosmetic. Windows users will find that their version of program has been very obviously ported from a Mac original, and this may or may not be a psychological barrier. And the graphics generated are very simple and stark, with no bells and whistles: not in any way a functional problem (on the contrary, I find it refreshing), but possibly a shock to users accustomed to emphasis on presentation over content. Neither would stop me whole heartedly recommending Viz!on in any context where science is done and non-scientific staff are part of the workforce.