Comparisons, Part I
A sea of data is silent and still. To unlock its secrets, you need both talent and a systematic approach. And most of all, you need to know which tools are available.
Each analysis is a comparison at heart. The possibilities today are not endless. Zelazny* differentiates among five basic comparisons: component, item, time series, frequency distribution and correlation.
Component comparisons show the proportions of individual elements to a total (e.g. the percentage of brands F, G and H sold in the regions North and South).
Item comparisons rank a select group of objects. Are all the same or is one greater or smaller, better or worse than the rest?
Time series comparisons demonstrate changes in objects over a period of time such as growth, stagnation or decline over the course of weeks, months or years.
Frequency distribution comparisons show how often certain objects occur in different consecutive intervals (e.g. income groups). The largest and smallest groups and indices such as average, median and range help in explaining the distribution of values.
Correlation comparisons show if the relationship between two variables is in line with expectations, such as a profit increase with a revenue increase.
Which comparison should you use when? The keywords of your question usually point you in the right direction. “Which percentage of sales comes from…?” is, for example, an indication for a component comparison.
The five basic comparisons cover the most commonly used business charts. A connoisseur of refined data analysis, however, would add clustering, shopping cart, geo analysis or other sophisticated comparisons to his toolkit. For more on intellectual tools for data analysis – stay tuned…
*Zelazny, Gene: Say it with charts – The Executives Guide to Visual Communication, McGraw-Hill, 2001, 4th edition.
Hey Dr Bissantz,
I have really enjoyed finding your blog (and the dog-blog of your friend Bella). I have been a Tufte fan myself for some years and its great to see that now there are software tools available for implementing some of his ideas. This posting prompted me to reply as I have been pondering on precisely this question but in respect to patent analysis. “How can we make good quality evidence based strategic decisions about our IP portfolio and potential patents?”
One of the better things I have seen in patent analysis so far is a visualisation by Aureka (http://scientific.thomson.com/products/aureka/) – yet it is still only a pretty picture and the analytical depth is not great. I would be really interested in hearing your views on who in the patent analysis software field is putting Tuftian principles in action.
In the meantime I have a masters student beginning in September at the Centre for Intellectual property studies in Gothenberg (http://www.cip.chalmers.se/) who will make a start on precisely this question. If you are interested I’ll keep you posted.
All best wishes
Wednesday, June 6th, 2007, 7:45 pmMatt (Netherlands/UK)