Reductionism vs. Holism in Data Visualization
Post date: Jun 04, 2010 8:45:25 AM
ho·lism [hoh-liz-uhm]
–noun Philosophy.
the theory that whole entities, as fundamental components of reality, have an existence other than as the mere sum of their parts.
Reductionism:
re·duc·tion·ism [ri-duhk-shuh-niz-uhm]
–noun
the theory that every complex phenomenon, esp. in biology or psychology, can be explained by analyzing the simplest, most basic physical mechanisms that are in operation during the phenomenon.
After reading Mike Loukides article on data science, I just found a very fitting description of statistics vs. visualization. Wherein statistics one might reduce the entire dataset to on single number indicating the significance of a hypothesis, the(first) approach of visualization is to show everything. What the user might find after looking at all the samples in e.g. a scatterplot will be different than the "mere sum of their parts" or mean value.
Holism:
A good example of this difference is e.g. Anscombe's quartet, which consists of four datasets with eleven data-samples. Each of these four different datasets, all have the exact same mean, variance and linear regression. However when displaying them as scatterplots, the difference is obvious.