Dear Potential Contributors to The American Statistician (TAS),
I greatly appreciate your interest in TAS. Each editor likes to place their own “stamp” on a journal. What follows are some of my thoughts concerning goals for TAS. Some of these appear in my inaugural editorial in Feb 06, see here. What follows is an expanded version.
General Goals:
In a nutshell, I have one thing I would like to do as editor of The American Statistician:
Promote the Discipline of Statistics
I feel that TAS, because of its “general” nature, and because it is a major journal of the American Statistical Association (ASA), is the one journal that is best suited to achieve this goal.
So, when deciding upon contributing a paper, I would like you all to keep in mind a thought: Will this paper promote the discipline of statistics?
But you may ask,
“To whom am I
promoting the discipline of statistics? Who is the target market?”
Of course, we want to satisfy our main existing constituency, namely ASA members who call themselves statisticians and have done so for a few years. However, I think we can easily satisfy this constituency as well or better than we currently do, while at the same time drawing a wider circle. I would specifically like to draw in those in the “massive fringe” – those who are teetering on the edge of ASA membership. This market is huge – several times the size of the existing ASA membership. These people are working in a wide variety of fields, acting essentially as statisticians, even teaching statistics, but are not ASA members. This is the group I would like to think about when deciding upon the suitability of papers for TAS. I want this group to read our papers, to cite them, to put them into action, to tell other people about the cool paper they read in TAS, so that others will read the papers as well, thereby growing our reputation and influence.
TAS has a lot to offer this “massive fringe” group, in terms of education about “standard” statistics. But what might not be so obvious to many statisticians is that the “massive fringe” has a lot to offer us as well. These people are scientists with discipline-specific knowledge that statisticians can use profitably. They are often applications-oriented to be sure, but some are highly technical in other fields (pure mathematicians, computer scientists, financial analysts, physicists) with fresh and useful technical approaches and insights that statisticians might not typically consider.
So, while I would not only like to publish educational, interesting and insightful papers about statistics that might appeal to this “massive fringe” group, I would also like to publish papers from this fringe group that will be educational, interesting and insightful to “mainstream” statisticians.
And then you may ask,
“But what is
‘Statistics’?”
Glad you asked!
In the spirit of promoting statistics more broadly, I take a broad definition of statistics. Those on the outside, e.g., some in the data mining community, have defined our field for us as something very narrow, having to do only with rigid hypothesis and modeling structures, divorced from subject areas.
Here is my definition of ‘Statistics’:
Statistics is the science
of collecting, summarizing, learning from, and making decisions from data.
As such, it covers just about every discipline, especially if ‘data’ is more broadly defined as ‘information’, in which case it includes text and even art and music. I wish to convey this broad sense of ‘statistics’ in the pages of TAS.
On Novelty:
I don’t think that “advances” are necessary, or even desired at all in some cases. Reviews and synthesis, fresh new looks (which indeed might be called ‘advances’), repackaging of existing materials in less accessible statistics or scientific literature, all are good. Papers that are strictly methodological advances should be sent to J. Amer. Stat. Assoc., J. Stat. Plann. Inf., Comm. Stat., etc. There are plenty of outlets for those types of papers.
As an analogy, Science (the flagship journal of the American Association for the Advancement of Science) publishes both technical articles and “user-friendly” summaries of some of the more interesting technical articles in the same journal edition. Personally, I rarely read the more technical articles, but I do read the summaries quite often. The writing style and the general approach of the summaries is an interesting and useful model for what I would like to accomplish with TAS, although our papers will generally be somewhat more technical than those in Science.
On Mathematics and Technical Details:
The preceding discussion might lead you to think that I am aiming to “dumb down” the journal. Let me assure you that nothing could be further from the truth! If anything, I would like to raise the theory bar, but only in a more focused, judiciously chosen way. Let’s have some serious real analysis-based probability theory, Hilbert spaces, information theory, etc. Our goal should be to popularize technical tools used by statisticians, as well as to popularize technical tools that statisticians do not typically use, but might adopt. We can do this by showing how such tools clarify, condense, and demonstrate elegant, inescapable truths laid bare, stripped of all annoying clutter. I would be happy to publish “cute” probability papers along these lines, as well as “hard” theory that supports even the most applied paper.
Results should be transparent, but NOT dumbed down. Sometimes the most elegant approach is the most transparent. To pick a stupidly simple example, use a matrix, not the written-out array, to clarify exposition.
However, papers should generally not have a lot of theorems and proofs, and more often none at all. Further, the heavy theory, if used at all, should be used extremely judiciously, always with the intent to clarify as I stated above, and never with the intent to impress readers with the authors’ technical expertise.
Additionally, the development of the material should be reasonably accessible despite any heavy theory.
On Controversial
Papers:
Bring ‘em on!
When I was a newly minted Ph.D. statistician, I joined ASA and started receiving JASA, TAS, and a few other journals. I learned much more about statistics in the broad sense from TAS, and I especially learned from the contentious articles, including those about Bayesian statistics and regression diagnostics for multicollinearity, for examples. The arguments back and forth made me think and inspired me to learn more than I otherwise would have.
I think controversy can be very stimulating. Sometimes you have to push the envelope a little farther than you might personally believe, just to get people to think. I suspect that W. Edwards Deming might have adopted that strategy. He didn’t *really* believe all that stuff, did he? Well, maybe he did, maybe he didn’t. Either way, he got us all to think, with his controversial, in-your-face points of view, and we are better off because of it.
On Scholarship:
It is essential that a careful review of the literature be conducted, and the contents of those papers should be integrated in a non-trivial way. That does not mean that there should be an excessive number of citations, but relevant ones must be cited. An analysis of papers accepted for a certain prestigious conference (not a statistics conference) showed that “# of citations” was one of the most significant predictors of acceptance - even more significant than peer review reports! This is a bad trend! Obviously, the right papers need to be cited, but excessive citations and/or gratuitous self-citations should be avoided.
Here are the kinds of citations that are important, in my view:
1. Cites to articles published in TAS. I want to promote TAS, to let people know what kinds of papers are there, and hopefully to get our papers cited outside the statistics profession.
2. Current cites. Current citations suggest that the topic is “hot” in some sense. Not that we should ignore the great, older literature, not at all, eg Jeffreys’ classic book.
3. Good cites outside of the statistical mainstream. In the interest of attracting interest from outside disciplines, we should look to include literature from outside disciplines, judiciously of course.
Glasnost:
Meaning “openness”, I believe, in Russian. Let’s make sure everything is above board. Data better be accessible, for example.
As another example of “glasnost”, I would like to publish what has been found useful in “hot” applications like credit scoring, anti-terrorism, fraud detection, cryptography, etc. Some of this information is hard to get, people like to keep it close to the vest, and in some cases there may even be national security concerns. But anything we can legitimately get on “hot” topics like these will be viewed quite favorably. Again – think promotion, promotion, promotion!
The Big Picture:
Maximize Benefit
Often, we statisticians get bogged down in minutia that is not particularly important in the grand scheme of things. I would like us to start “getting real” in TAS. By this, I mean I would like to encourage statistical reasoning that profitably contributes to the “bottom line,” whatever that may be. It may be in terms of real profits, as in a business setting, or it may be in terms of reduced time from hypothesis to theory in scientific enterprises, or it may be in terms of promoting human, animal, or planetary well-being. If we are to appeal to the aforementioned “massive fringe” group, we really need to aim in this direction. Papers that show beneficial uses of statistics in these regards will be considered favorably.
Bad Statistics in
Practice:
As I said, I don’t mind a little controversy. One thing I would like to see more of is some critical appraisal of statistical methodologies that are commonly used in other disciplines, with suggestions for improvement, if warranted. Of course, any review of such critical appraisals should include opinions from the inside of the disciplines that are critically appraised, or perhaps there could be a discussant from inside the discipline. This kind of paper would be a clear way to catch the interest of some of those in the target market who are in the “massive fringe”.
Write Gooder than Me:
Many of us are techno-geeks with backwards social, verbal and written skills. Some of us bathe infrequently and attend Star Trek conventions. Let’s face it: We are no Mark Twains! But doggone it, we can produce well-written articles, if we try. I want people to read TAS papers. So they had better be readable!!!!!
Rating System:
In addition to the usual comments, associate editors and reviewers will “grade”
papers with 1-5 scales on the following four dimensions:
Is the paper of general interest?
Will this paper have an impact?
Is the paper written well?
Is there adequate overview?
Please keep these dimensions in mind when submitting papers.
Final Words:
This overview is not meant to supplant the fine papers
written by former editors Jim Albert and Lynn Stokes; see http://westfall.ba.ttu.edu/JimAlbert.htm and http://westfall.ba.ttu.edu/LynneStokes.htm. In particular, Jim’s article has great
reasons for when a paper should be rejected, and
Thank you again for your interest in The American Statistician,
Peter Westfall
Editor-Elect, TAS