Dear Colleagues:
I promised in Tuesday's post that my next post would be the last in my planned set of three. I appreciate the comments that others have sent in response, and I hope their authors will forgive me if I don't take up bandwidth replying to each. Rather, I hope the conversation will continue without me, or if not, will die a natural death.
That said, here is a delayed third post, revised thanks to the tact and thoughtfulness of our Executive Director, Ron Wasserstein. My hope is that in this last shot at your attention, you will indulge me in some undisguised but heartfelt preaching:
· Not one of the ideas in my two previous posts was original with me.
· To the extent that I may have engaged your attention, it was not from offering new ideas. Rather, it was from using words to call attention to the important ideas of others.
· One can stand on the shoulders of giants and at best see no farther than they did. But if you work at crafting the message, you can help ensure that the voices of prophets may reach more people than otherwise.
· If you want your work with data to make a difference, devote time and effort to choosing the words and pictures you use to present your evidence and conclusions. If you teach or supervise, seek to reward those – they who learn from you, and they who report to you -- when they spend time crafting their message.
The moral of my homily: How you craft the message often has more impact than the message itself.
Those of us who teach statistics have learned the importance and impact of a strategy that I will simplify as a two-stage process, a strategy that I learned from David Moore, one of my handful of heroes: (1) First, identify the essential ideas, and then (2) compress/express those ideas into/as what David has called "slogans." A generation of students has learned basic concepts of statistics from Moore's slogans like "Plot, shape, center, spread" to guide thinking about a univariate data distribution, and the "68, 95, 99.7 percent rule" to reduce the normal distribution to words. Freedman, Pisani, and Purves used the metaphor of random draws from a box of numbered tickets to make concrete the idea of a probability model, and then condensed the ineluctable dependence of inference on probability into four pithy words, "No box, no inference." In their book, of course, the slogan should have appeared in a box. Box himself gave us the memorable slogan, "All models are wrong; some are useful."
My main point is that just as a lot of data analysis is based on skills that can be taught and learned, so is effective exposition. Think back to the last "report" you were expected to read. Is it easy to recall the main points? The answer, yes or no, depends not so much on the quality of the data, the effort, and the thinking that went into the report, nor on your own dutiful diligence in reading the report, but rather, and mainly, on whether the people who wrote the report had learned and practiced the skills of how to use words and pictures, first, to claim attention, and second, to claim retention: to deliver a message that sticks in the mind.
When it comes to failure to make the message triumph, who owns the loss: the general of the battle or the nail on shoe of the horse? The time and effort spent gathering data and thinking about its message is the main battle, but the missing use of effective prose is the lost nail, lost shoe, and lost horse, for want of which the battle of evidence may be lost.
Words matter, but we should not shrug off the way we use them as just the poker hand we've been dealt by fortune. A good poker player learns to do well regardless of the hands dealt. Using words well is much more a skill you can learn than a talent you inherit. I urge any of you who have stayed with my rant so far to seek out, read, and take to heart David Moore's acceptance speech on being chosen for a teaching award by the Mathematical Association of America. Moore argues that teaching is not a talent you are born with, but a craft you can learn. I'm sure he would agree that, like teaching, any form of communication is a craft that you can learn, not a talent you are born with or without.
What should we do to improve the way we members of ASA communicate about the importance of data? Here are three suggestions, one short term, one long term, and a third, in between and more tentative.
1. Short term. The American Statistical Association is hosting a Symposium on Statistical Inference, October 11-13, 2017, at the Hyatt Regency Bethesda (Maryland). The tagline for the symposium is "Scientific Method for the 21st Century: A World Beyond p < 0.05."
One of the goals is "Making evidence available and accessible." Consider attending. Post your own experiences: stories are persuasive.
2. Long term. I urge every ASA member to join me in thinking about the connections between (a) our goal to make the importance of learning from data prominent in the popular press, and (b) the skills we can teach and learn to make this happen.
3. Intermediate: We might all think together about possible ASA initiative proposals for next year based on the theme, "How to use words and pictures to raise the profile of data in the popular press." As I see it, this is the most difficult of my three suggestions. There's a hard deadline, and a soft goal. I won't try to count the flowers, but a thousand would be nice.
In conclusion, I repeat my debt to Ron Wasserstein, whose tact and thoughtful comments on a preliminary draft of this post have, I hope, spared me the embarrassment of charging into the fray riding a horse whose shoes might have been a few nails short.
Thanks, Dr. W.
But, Ron, please know also, that should I be pilloried, you, as the last person to offer me advice, will be the first person I blame.
George