I've been using this blog as a personal public notepad of graphics related things (with an occasional comment here and there). I looked long and hard for a blog devoted to such items of interest, but I came up fairly empty-handed in my quest, so I decided to just take a note here and there subject to waxing and waning time and inclination. (I do recommend an RSS feed on the graphics tag at del.icio.us.)
I'm going to deviate from the occasional link-of-note post and make a book recommendation, a recommendation for an old favorite. The problem with making book recommendations is that there's always going to be someone who says "Yeah, well, duh!" somewhere, somehow, even if not in the space of comments somewhere out in the ether, I'm sure. (Not that I really have to worry too much about comments--anything outside the realms of snide remarks and spam will be quite welcome.)
In spite of the obligatory yeah-well-duh companions with book recommendations, there's always a first time for everyone with every book, and this post is meant for those who are unfamiliar with a little book titled "How to Lie with Statistics
," which is a great short read, and it doesn't require anything near a degree in mathematics.
One reason I'm recommending this book is that I haven't seen it recommended very often (and certainly not enough), and I don't want great books to be forgotten. A late mentor of mine, a public policy professor, used it as one of the references in classes he taught. I also ran into it on Jim Blinn's page
in a section titled "Really Good Books That Have Changed My Life." I can't think of seeing it reference any where else for some time.
It's an old book now, its 50th anniversary was last year. Under the guise of teaching you how to lie with statistics, author Darrell Huff does a wonderful job of showing how fact and figures can be misused and abused as they often are both intentionally and unintentionally. Time and time, again I've found myself speaking with well-educated adults who have fallen into some of the classic statistical traps described in the 142 pages of this book.
I realize that if this were a really decent review, I'd devote much more time and space to the actual contents of the book. The table of contents and excerpts are available at the Amazon link above. My chief concern is doing my part to raise awareness and currency, but I will give an example.
Once I attended a market research presentation. A major finding of the research was that 50% of a particular company's sales were made to corporate customers. This struck me as a surprisingly high number, so I asked how the number was determined. When it was revealed that the number was derived by analyzing returned registration cards, it was clear that the entire study relied on the highly questionable assumption that home users are as likely to fill out a software registration card as corporations with organized I.T. departments.
Getting a statistically valid sample is tricky business. It's often very hard to do, and I've frequently seen some egregious errors made by people who should know better. Huff opens with an example that includes a number of potential sampling errors factoring into the formulation of a reported average salary for Harvard graduates; he notes how salaries tend to be overreported, the most prosperous people are often the easiest to track down, etc.
In closing, I'd like to say that even though my old mentor used it as a reference for college coursework, I think this book is accessible enough for high school students--if not generally, at least in advanced placement classes. If you're a high school teacher in an appropriate discipline, I feel you could do your students a great service by using it as a reference given the extent to which facts and figures are presented in the mainstream media. We need critical thinkers in today's society, perhaps more now than ever before.