Graphical metaphor for dressing elegantly

[Metcalfe]

Professor Emeritus Edward Tufte asserts that Minard, a French engineer, drew the finest statistical graphic in 1861 when he described the history of Napolian's Army during its Russian Campaign of 1812.* Tufte never says "elegant" but I think his remarks are consistent with Balzac's definition of elegance:

"Minard's graphic tells a rich, coherent story with its multivariate data...

[Minard's graphic illustrates] how multivariate complexity can be subtly integrated into graphical architecture, integrated so gently and unobtrusively that viewers are hardly aware that they are looking into a world of four or five demensions."*

Here's the graph: http://www.edwardtufte.com/tufte/posters

I think Tufte's observations about integrating complexity gently and unobtrusively also apply to dressing elegantly. The Duke of Windsor dressed with more variables (e.g., patterns) than anyone during his era, but he integrated them gently--usually.

Minard's graph was drawn by hand as if it were a bespoke suit. He designed the graph for Napolian's Army's story with no expectation that he would reuse the pattern. Minard's graph is the opposite of an Excel bar chart wizard--the statistical equivalent of ready-to-wear.

A bespoke graph assists with telling a complex story; a bespoke suit assists with listening to a complex story. That's why bespoke suits endure for professionals whose advice requires listening unobtrusively. Bespoke suits absorb stories, and the more elegantly we wear them, the more stories we hear.

Metcalfe

* Tufte ER. The Visual Display of Quantitative Information (2nd edition). Cheshire, CT: Graphics Press; 2001. Page 40.

[SouthPender]

As a person who works with multivariate data every day, I'm interested in this notion that the many variables that go into a bespoke garment could be graphically depicted. However, I'm at a complete loss as to how this could be accomplished. The fact that a graph of a phenomenon is drawn by hand (in this case because no other method was available at that time), doesn't suggest, to me at least, that it would describe something like a bespoke suit any better than any other depiction of multivariate phenomena. In addition, when you say that the Duke of Windsor "dressed with more variables" than others, this is hard to understand. The same variables underlie all bespoke garments (lapel shape, shoulder treatment, etc., etc.), and come into play regardless of who commissions (or makes) the garment. Maybe what you were saying is that the Duke of Windsor demonstrated more variation in his garments (each based on different variations of the same underlying multiple variables) than anyone else. But getting back to your original point, I'd be very interested in hearing how the bespoke process could be analyzed via modern multivariate methods, including a graphic depiction of the process.

[Metcalfe]

Dear SouthPender,

I apologize for the confusion. I wasn't suggesting that we graph bespoke suits; rather, I was suggesting what we might learn about wearing suits from the principles of multivariate graphing. It's an indirect response to Mr. Alden's post on elegance in which Balzac asks science and other disciplines to inform our discussion of elegance.

It is true that suits entail core variables: shirt, tie, suiting. The typical statistical graph has its core variables: dependent variable, predictor, covariables. Tufte brings Minard's graph to our attention because it has more variables than the typical statistical graph and it integrates them gently. We're not aware that Minard's graph as five variables; we just see the engaging story of Napoleon's Army. The same is true of the best art, notably opera: We enjoy a well-integrated story when no single variable (e.g., the set) stands apart from the overall story.

I think the same is true of elegant dressing, although you are correct to say that I mean both variables and variation within variables. The Duke of Windsor (and Prince Charles) can add variables and variation effortlessly: pattern to the shirt (variation), pattern to the suit (variation), a pocket square (new variable), a buttonhole (a new variable), cufflinks (a new variable), etc. Prince Charles--or one of his four valets--designs an ensemble so gracious that we see an overall effect and lose track of how many variables and how much variation are working for our pleasure. Thank you for taking the time to correct the metaphor.

My second point is more controversial and I should have marked it accordingly. An elegant statistical graph presents a story; a suit can help present a story, too (e.g., a lecture or presentation). Balzac and Lord Chesterfield would press us harder: A suit functions at its most elegant when it receives information, when it helps the gentleman listen to stories. Members of the professional and governing classes wear bespoke suits because information flows both ways, primarily toward them. Ministers (civil and religious), doctors, attorneys, consultants, and professors must listen before advising their constituents, parishioners, patients, clients, and students. A bespoke suit worn elegantly draws people toward the advisor and elicits stories. Lord Chesterfield teaches his son these principles of how to gather foreign intelligence graciously (and how to charm women by listening attentively).

Does the graphing metaphor work for dressing elegantly? What other metaphors can we develop to appreciate the nuances of dressing elegantly?

In my original post I claimed that Tufte never wrote "elegant" but I was wrong (see pages 176 and 177 of The Visual Display of Quantitative Information):

"Minard's refined use of color contrasts with the brutal tones often seen in current-day graphics. What makes for such graphical elegance?...Graphical elegance is often found in simplicity of design and complexity of data."

Metcalfe

[SouthPender]

Metcalfe, thank you for your clear and reasoned response. This idea interests me. In my work with multivariate data sets, our goal is usually to combine multiple variables (usually predictor variables, but also dependent variables, and covariates) into maximally-effective linear composites so as to achieve some outcome of importance--like predicting a patient's likelihood of benefitting from treatment, or a company's likelihood of succeeding in the marketplace. The models we use are compensatory, in the sense that low values on one variable can be compensated for by high values on another. This kind of model would not work in developing, say, a multiple regression equation for the ideal bespoke suit. Here you would need a model that kept values on the variables within certain ranges to avoid the problem you note of a single variable standing apart from the rest and, perhaps, dominating the equation, with the result being a very bad suit. This can be accomplished with conjunctive models, but your idea of eliminating the possibility by smoothing the phenomenon via graphic means is quite intriguing. I particularly like the idea of the outcome having stories to tell us. I must give this some thought with respect to the usual--far more mundane--applications of these methods with which I deal!