John Tukey, Exploratory Data Analysis charts

Statistics are a hard sell for the non-initiated, among who I count myself.
However, having been trained in visual graphics, I think I could appreciate, although a little, the effort made to compress meaning into these geometrical shapes. The same encoding taht make the charts and graphic so hermetic to the lay people like me, who just recently could appreciate the difference between logarithmic and linear scale in the confection of a graphic.

There is a method to the madness, somebody would say, and it is in the interpretation of this method wee the magic process of decodification and interpretation happens.
These are some screenshots form the charts John Tukey used in a serie of talks about the work devloped in his book Exploratory Data Analysis (1977).
Among other things he proposes an improved version of the boxplot (box and whiskers) for the visual analysis of the characteristics of a dataset (highest/lowest value, outliers, median, etc).
I liked the rough aspect of the pre-internet typography and photocopy of the pdf I found, it gives, for all the statistics glamour, acertain sense that, this, is the work of a person, who like everyone else, manage a certain degree of uncertainty in his life.
Because sometimes the greatest thing you can admire in most sciences is this prometeic effort to grapple with the world, many times obscures by intradisciplinary jargon and referencing.
But as much as these screenshots can show, a group of graphic drawn on paper, there is much more to the work of Tukey, like I discover when seeing his video demo showing the PRIM-9, which you can see here: http://stat-graphics.org/movies/prim9.html, a data processor of sorts that prefigurated the Tableau of our days. In the video  which I will link below, you can see how Tukey manage and mmanipulate a set of datapoints in the monochrome screen.
 And this was in the seventies!. And I also didn't know that Tukey and Ascombe were related trough the family of their wives! But that is another subject.

.






Comments