My former department chair, Xiao-Li Meng, has published an excellent article on the emergent role of statistics and the challenge of teaching the science to non-statisticians. He addresses the negative perception of the field, often ingrained by a poor high school experience and summed up in a dismissive scoff that "the best speaker in statistics" is hardly an accolade at all.
Ultimately, he views statisticians as quantitative authorities:
We statisticians, as a police of science (a label some dislike but I am proud of; see the next section), have the fundamental duty of helping others to engage in statistical thinking as a necessary step of scientific inquiry and evidence-based policy formulation. In order to truly fulfill this task, we must constantly firm up and deepen our own foundation, and resist the temptation of competing for “methods and results” without pondering deeply whether we are helping others or actually harming them by effectively encouraging more false discoveries or misguided policies. Otherwise, we indeed can lose our identity, no matter how much we are desired or feared now.
I think the title is appropriate but limiting; it isn't the statistician as a person who is a police officer as much as it is the proper application of the field itself that is designed to eliminate poor analyses. Statisticians are merely the people trained with such knowledge, and there is nothing preventing a statistician from performing analyses of his own so long as he is able to properly moderate his own work.
Professor Meng's most salient point in my mind comes near the end, when he calls for more attention on the quality of teaching, particularly in introductory classes:
With their potential impact in mind, it is easy to see the necessity of having the most qualified teachers for these introductory courses, just as for more advanced ones. And if I had to make a choice (and sometimes I do as a department chair), I surely will give the general introductory courses the highest priority for a very simple and practical reason. If an advanced course is sabotaged by bad teaching, the chances are that it will only affect a relatively small number of students, most of whom would have, or already have had, another chance to study statistics and to be convinced of our beloved subject’s beauty and importance.
In sharp contrast, if a general introductory course is badly taught, it often will affect hundreds, or even thousands, of students, and the vast majority of them will never take another statistical course, even if some of them initially had some curiosity or interest in statistics. This is very much like a badly taught AP statistics course that can do more harm than help, permanently turning away many of its students, as all they saw was “Oh, this is what statistics is about—boy, am I glad that there are many more interesting and relevant subjects in college than this!” Indeed, among the Harvard undergraduates I asked, the most frequent reason for not considering a statistical major was a “turn-off” experience from an AP statistics course.
I think this is a great paper with a teaching message I firmly believe in. It's a short read with absolutely no technical nonsense - its tone is actually almost colloquial.
(via Andrew Gelman)