Anscombe thought traditional statistical schooling under appreciated the role that graphs play in enhancing the understanding of data. He believed three ideas permeated academia:
- Numerical calculations are exact, but graphs are rough.
- For any particular kind of data there is just one set of calculations constituting a correct statistical analysis.
- Performing intricate calculations is virtuous, whereas actually looking at the data is cheating.
There is nothing inherently “gross” about graphical analysis. Graphs can be precise and granular, like statistical models. Graphs provide another window of insight. Anscombe illustrated how valuable graphs can be for understanding data. I’ve re-created Anscombe’s Quartet in the following Tableau dashboard:
Anscombe’s point is, despite having nearly identical variance and correlation figures, the data sets are distinctly different and that the use of graphs helps you see and understand those differences. Try changing the simple statistics displayed at the bottom of the dashboard by selecting each of the data series. The calculated values are nearly identical.
Seeing the data graphically is helpful to guide additional statistical analysis.