Stats Made Easy

Practical Tools for Effective Experimentation

Sunday, November 25, 2007

Statisticks (sic)

This Fall at a statistical conference I attended a stimulating talk on improving communications. The speaker recommended a book called Made to Stick. The authors, brothers Chip and Don Heath, provide six principles for making messages memorable: simplicity, unexpectedness, concreteness, credibility, emotions and stories (SUCCESs). This article by News and World Report provides more details on the tricks to Making it Stick.

Our company’s mission statement, Statistics Made Easy®, has proven to be very sticky. It is simple, very unexpected, concrete and emotional (statistics are scary!). However, the never-ending challenge for Stat-Ease is to make its slogan credible.

The Heaths provide an enlightening discussion on how to convey statistics better. They provide, for example, the challenge of being the publicity director for a group committed to saving sharks. A statistician might say that on average only 0.4 Americans die per year from shark attacks (that seems very macabre!), whereas in 2000 alone, 12 died while swimming off lifeguard-protected beaches. These stats do not stick nearly as well as this alternative: Which wild animal is more like to kill you – a shark or a deer? Answer: A deer is 300 times more likely to be the death of you. Oh dear!

“Statistics are rarely meaningful in and of themselves…It’s more important for people to remember the relationship than the number.”
– Chip and Dan Heath

Sunday, November 18, 2007

A laugher of a curve offers a lesson on leverage

A former colleague of mine, Jim Mork, sent me the ‘heads up’ to this blog by Mark Chu-Carroll, a PhD computer scientist who works for Google. He and others* mocked the Wall Street Journal (WSJ) for a poorly fitted curve of tax revenues versus percentage of gross domestic product (GDP) by country. They claim that it confirms the Laffer curve, which quantifies the diminishing returns that a government gains by going beyond a certain level of taxation.

"It should be known that at the beginning of the dynasty, taxation yields a large revenue from small assessments. At the end of the dynasty, taxation yields a small revenue from large assessments."
-- Ibn Khaldun (14th century)

The WSJ curve starts from the origin at 0 percent corporate taxes for United Arab Emirates (UAE), which obviously contributes 0 to their treasury (presumably their money comes from a general sheikh down). The WSJ graphic designer then took the curve all the way up to Norway at 10 percent on the y (upper) axis, ignoring the dozen countries below this in the mid-range of rate of taxation on corporations. This is a case or providing one point (Norway) an extreme leverage in curve-fitting.

I searched out comparable Organisation for Economic Co-operation and Development (OECD) figures from a few years earlier (2002) than those upon which WSJ based their analysis. My curve-fit, done in with Design-Expert® software, does show a characteristic Laffer peak at 29 percent corporate rate, thus supporting the argument for reducing our USA rate of 35 percent. However, as one can see from the very broad 95 percent confidence band, this finding is not one that government officials ought bank on.

*Kevin Drum of the Washington Monthly

Monday, November 12, 2007

Why experiment?

Last week I presented a one-day workshop on design of experiments (DOE) to a diverse group of Six Sigma black-belt candidates at The Ohio State University’s Fisher College of Business. Having done this many times before, I’ve grown accustomed to some ‘aliens-in-the-headlights’ looks when I put these students to the test of actually doing an in-class experiment on paper helicopters. Once an accountant pulled me aside to say that something must be wrong with his team’s experiment because the results varied – some ‘copters clearly flew better than others. I realized then that the whole concept must be unclear to those with no background in science or engineering.

Therefore, for last week’s presentation I prepared an ‘icebreaker’ on why experimentation may be needed. I wrote these three words on the marker-board before class: “Desperation,” “Domination,” and “Doubt.”

Desperation from the Great Depression motivated Franklin Delano Roosevelt to say “Try something. If it works, keep doing it. If it doesn't, try something else. But above all, try something.”

Domination is what Google gains by their practice of experimenting “quickly and often” as noted in a blog by John Hunter (son of DOE guru and ASQC Statistics Division founder Bill Hunter). What if you could test not just one factor at a time, but many? You can via two-level factorial designs!

Doubt is engraved in the lobby floor of Fisher College’s Pfahl Hall where I taught my class. It says “Who doubts nothing, knows nothing.”

I hope these three reasons sufficed to inspire this crop of black belt candidates to give experimentation a try.