Truth, damn truth, and statistics

Practitioners and statisticians are often forced to rely on the popular belief that statistics teach how to deal with data. Those who mistakenly believe that statistics are purely a branch of mathematics (and therefore algorithmic) often view the use of judgment [decision making] in statistics as proof that we are indeed manipulating our results. In an effort to give correct formulas and definitions, we may not notice the important role that judgments play. We must teach our students that they are personally responsible for making statistical decisions. But we must also offer some guidance for [building] statistical decision rules. Such leadership requires that we recognize the role of ethics in statistics. The principle that determines these rules should be an honest search for the truth about the world, and the principle of searching for such truth should be central to statistics courses. The remark attributed to Disraelia will often be applied with a certain degree of justice and power: "There are three kinds of lies: lies, blatant lies, and statistics" (Mark Twain, Mark). This may be my least favorite quote about statistics. But I would like to dwell on what lies at its core, and on the joyful willingness of many who know nothing about statistics at all, to quote it as if it justifies their low opinion of discipline. This quote has penetrated discussions across many disciplines. They will probably tell you that you are stupid enough to admit to a decent company that you teach statistics. Nigel Rees1, b claims that this is the only frequently used remark in the British media2. A Google book search [by meme] "lies, blatant lies and statistics" shows 495 books, and a general Google search finds "about 207,000" citations. A small (non-random) sample of these links indicates that the majority are intended to offer dishonest manipulation and interpretation.