Skip to content

Could I use a statistics coach?

In a thought-provoking article subtitled “Top athletes and singers have coaches. Should you?,” surgeon/journalist Atul Gawande describes how, even after eight years and more than two thousand operations, he benefited from coaching (from a retired surgeon), just as pro athletes and accomplished musicians do. He then talks about proposals to institute coaching for teachers to help them perform better.

This all makes sense to me—except that I’m a little worried about expansion of the teacher coaching program. I can imagine it could work pretty well for teachers who are motivated to be coached—for example, I think I would get a lot out of it—but I’m afraid that if teacher coaching became a big business, it would get taken over by McKinsey-style scam artists.

But could I use a coach?

First, let me get rid of the easy questions.

1. Yes, I could use a squash coach. I enjoy squash and play when I can, but I’m terrible at it. I’m sure a coach would help. On the other hand, I’m happy at my current crappy level so I won’t bother. (Paradoxically, I think if I ever play enough that I actually start to get better, I’d probably want to get better still, and then maybe I would want that coach.)

2. Yes, I could use a teaching coach. I have lots of flaws as a teacher and I’m sure a coach would help. I’m not so sure that all or even most college teachers would benefit from coaching. Not because they’re so good, but because I think you have to want to improve, and I think lots of teachers would prefer to just not hear about their own flaws. (Gawande talks about that with surgeons.)

And now to the tough question:

3. Could I use a statistics coach? Gawande reports that top musicians such as Itzhak Perlman (as well as, of course, top athletes of all sorts) benefit from coaching. But I’m not clear that coaching would benefit my performance as a statistician. Coaching might help me with my management skills and other peripheral items, but I don’t know that someone could watch me do statistics and point out areas of improvement, in the way that Perlman’s coach can. I mean, what would there be to watch, exactly?

In this way, then, I think Gawande’s article is missing something. Let me put it another way: Gawande is a fine surgeon (at least, from the evidence he presents in his article) and it seems that coaching makes him even better. But, despite “surgeon” being a central part of his identity, it’s not what makes Gawande special. The United States is full of skilled surgeons making a million dollars a year (or whatever it is they make). What makes Gawande special is not that he’s one of that group; rather, he’s special because of his skills as a reporter and writer.

So the real question is: Is he getting coached on his reporting and writing? I think it’s possible. To be a good editor, you have to have some writing skills but you don’t have to be a top top class writer. In fact, as noted in this space on occasion, in my dream job (if I could not be a statistician), I’d be a Max-Perkins-style editor (if only such jobs still existed).

But I don’t know if this could work for statisticians (or for physicists or computer programmers or various other technical jobs). I’m sure I could benefit from advice—if I had Don Rubin or Xiao-Li Meng or Jennifer Hill on a string to answer my statistics questions at all time, I’d be in much better shape (this is not possible so I have partitioned off areas in my brain to simulate Rubin and Meng and Hill—it’s not as good as the real thing but it actually can be helpful, sort of like those old Windows emulators they used to have on Macs)—but that sort of advice and feedback seems a bit different from coaching, somehow.


  1. Gabriel Rega says:

    Isn’t peer review a bit like coaching in some sense? You show the reader not just what you’ve done but how you did it. Reviewers then come in and say what you missed, how you could have done better, ask why didn’t you follow some other proceedure, etc?

  2. Brent Buckner says:

    Maybe coaching could help with achieving/maintaining flow state for doing your statistics work.

  3. Erik Shilts says:

    I loved the article and I think you’re wrong on the need for coaches in technical sectors. Computer programmers constantly use coach-like feedback mechanisms to make themselves better. All tech companies have code reviews and there are a number of programs that facilitate reviewing code and commenting on it. Even putative code storage sites like Github offer the ability to comment on commits or certain lines of code and mechanisms to improve others code by submitting your own.

    And it’s more than just coding. I work on a statistics and data mining team at a rapidly growing software company and we utilize our weekly code reviews to provide feedback on statistical methods and suggest alternative analyses.

    These reviews don’t utilize coaches per se, but the mechanism is all the same.

  4. Jeremy Miles says:

    I’d love someone to sit next to me all the time and say “there’s a typo in that code, and you haven’t closed that bracket”.

  5. tc says:

    Maybe you’re at a point where you don’t need to worry about it anymore, but I sure could use an academia/publishing coach. In every field there are some guys who put out a ridiculous number of papers (e.g. Acemoglu in economics), maybe that’s because they’re just smarter, but some of that has to be due to a system that works for them.

    • C Ryan King says:

      I remember when Tyler VanderWeele was in my department; he was a productivity machine. It’s probably mostly self-discipline, but I can imagine him sitting down and just getting some of those things right the first time.

  6. Anyone who publishes in the New Yorker is essentially getting a kind of editing much akin to coaching! But I think you are missing something in your analysis. You and I, in our lives as senior academics, essentially _are_ coaches rather than _getting_ coaching – the work that we do with graduate students and even junior colleagues involves the sort of meta-intervention (on how to define projects, on the best ways of tackling different kinds of problem, on how to save time and trouble or at least to work with flawed existing materials and improve them) that is, essentially, what the term coaching means. If editing is coaching, then so is that sort of teaching (as distinct, perhaps, from undergraduate teaching, which involves less of this sort of meta-work on skills, approaches) – though it can’t be broken down as clearly, perhaps, because it is not an art in time, it’s not ‘live’.

    • Sebastian says:

      I thought about that – but what then to make of the Perlman example. In the world of Violinists, Perlman would certainly have the equivalent of an endowed chair at an ivy-type institution. (actually, I don’t even need the analogy – Perlman _does_ have an endowed chair at Juilliard). Same applies to Renee Flemming.
      So if he can use a coach, why not Andrew Gelman, Don Rubin, Gary King etc.?
      Although if you read what they say about their coaches, that really doesn’t sound so different from a colleague giving you feedback about your work.
      So maybe academics has had a system of “coaching” each other for ages, we just give it a different name?

      • Yes, giving talks and getting reader’s reports on articles and book manuscripts and having really sharp focused conversation now and again with colleagues would all seem to me to fall under the coaching rubric. There is more of this in, say, pure math than in literary studies, possibly because there is a stronger ‘practice’ element in math – for performers and practitioners, coaching is more essential because it is so useful in the matter of correcting habits (whereas the fact of being able to write drafts/revise means that we can do more of our own coaching when it comes to the written word/symbol – we refine it over time, with help, and then present a static final version in print, so that the analogy can only go so far).

  7. K? O'Rourke says:

    I am a big fan of this idea(having been involved in boxing where everyone literally pays for bad habits they allow to creep in).

    Essentially the idea scientific process is make one’s work available to your peers so that they can effectively observe, judge, and guide (i.e. I could not replicate).

    For statistics coaching, you likley would need to verbalize out your thinking process to make that accessible to a coach.

  8. akshay bhat says:

    I think he misses an important point, Coaches coach because they cant play! If coaches could play then we would have an hierarchical structure in every team, where as a person ages he gets more authority, and thats exactly what happens in surgical teams, where there is a chief surgeon at top and multiple residents working under him. The reason this does not happens in sports is because the ability (physical ability) peaks at a certain age, which is not true for other professions.

  9. John Mashey says:

    1) A personal coach spends dedicated time to help someone perform better, usually by being able to watch and give good feedback.
    Example: a Wii can be a balance coach, as it can give precise, immediate feedback (in a way no human can, although I’ve seen ski caoches who could detect small left-right differences pretty well.)

    2) Peer review isn’t that.
    In any case, by the time something gets to peer review, or even giving a talk, there may have been quite a bit of development, exploration of blind alleys, mistakes that got fixed.

    3) Hence, to really have a coach, one needs somebody that spends enough time to see these issues and point them out.
    For example: a coaching technique in R&D development is for everybody to draw a (logistic-looking) curve of when they think some project will be done, record those, and see what happens. Another is to spend some time thinking about the exploration of the design space, and see if anything can be learned for efforts that were fruitless, but should have been know as such earlier.
    As an analogy, think of Alpha-Beta Pruning of search trees in games.

    4) I’d guess that ion many research collaborations, people may well act somewhat as coaches for each other.

    • John Mashey says:

      Oops, and of course, in well-run R&D organizations, managers are expected to coach people often.
      Certainly, at Bell Labs, we did, including making sure that people got help from other disciplines as needed, including statistics.

      For statistics,see this which mentions Tukey, Kruskal, Chambers, Cleveland as a small sample. While such folks had their own research areas, it was hardly atypical for others to come looking for early help, rather than getting surprised with a negative review late in the process of submitting an external paper.

      if I understand aright, many universities have statistics-consulting services, like Stanford’s, which in some sense might be considered “drop-in coaching clinics,” obviously not services consumed by statistics experts. From afar, this sort of thing sounds like a valuable service. Maybe such deserves its won discussion thread: what different ways are these services organized? How well do they work?

      • K? O'Rourke says:

        I recall a conversation with a senior university administrator when they were a dean.

        All my falculty are world leaders at something (Ivey league, but some bragging here).

        But they are all bad at something – its my job to find a means for them to discover that and make needed improvements
        – without them noticing (if possible!)

  10. perhaps this discussion suggests that we ought to change the peer review process. instead of “coaching” only upon having a finished product, we could coach one another along the way. this would be a radical change towards open science, but perhaps for the better? perhaps an intermediate stage would be to develop our work openly amongst trusted colleagues, but not openly for all, somewhat like in sports: athletes get coached beside their peers, friends, teammates, other coaches, etc.

  11. Maybe? Coaching costs money -in some areas it pays for itself, that’s probably less clear in areas where an improved performance doesn’t lead to increased income to pay for the coach. I’m a fan of the concept of life coach, as low status as the title is and as full as hucksters as the profession may be.

  12. Andrew McDowell says:

    Coaching is often used in preparation for short bursts of concentrated activity – this covers sport, musical performance, surgery, military combat, and could cover lecturing. In this case, it seems worthwhile to spend time outside these few bursts of concentrated activity so as to perform as well as possible during them. Where this is not the case, time spent being coached must usually be subtracted from time spent working.

    Another reason for a successful researcher to seek training would be to prepare to apply skills from another discipline. A non-statistician might seek training in statistics. A statistician might seem training in an area to which they wish to apply their knowledge. It is possible, however, that such training would happen informally as part of collaborative working.

    Feynman recounts some not very productive adventures in biology in the chapter “A Map of the Cat?” in “Surely you’re joking, Mr.Feynman”.

    • Yes, on your second paragraph.

      This is exactly what I did. I am an astronomer, and I became interested in Bayesian approaches from reading UseNet and wondering about what it was all about; but did not have the background to use it in my field. By good luck I read an article by Jim Berger and Don Berry in American Scientist; I contacted Jim and this has resulted in some papers and a lot of learning on my part. A sabbatical made it possible for me to work with Jim and Peter Mueller at Duke in 1999. And the relationship continues.

      So this is a kind of coaching, or mentoring, that I wholeheartedly support. It’s good to teach old dogs new tricks.

  13. The most effective “coaching” I’ve found for programming is pair programming with someone else. Even if they’re more “junior” than you are, you can often learn something for them. And it’s amazing how much you can learn when pair programming with someone who knows a particular system or technique as well or better than you.

    I agree that getting reviews (code reviews, grant reviews, paper reviews, working with a co-author) is like being coached. Successful academics usually spend a substantial amount of time (implicitly or explicitly) asking each other how to do things, Andrew included.

    Gawande’s examples of coaching were all on the mechanics of surgery, not on the judgment involved. I’m not sure what the analogy is for a statistician. Being better at analysis? Programming? Modeling? Writing?

  14. K? O'Rourke says:

    Hmm.. running a blog often results in a lot of coaching – doesn’t it?

    As an aside heard another great Don berry quote the other day “Multiplicities humble us all!” (no date or reference given)

  15. yolio says:

    In academia, we call it mentoring. We could all benefit from good mentoring, but it is very hard to come by. And as you climb up the ladder, it gets really, really hard to come by good mentoring. I don’t think a good mentor needs to be a “better” statistician than yourself. In music and athletics there are plenty of examples of so-so musicians who shape up to be genius coaches. And brilliant athletes who can’t coach at all. The problem is that academics has a go-it-alone reward structure. Being a great coach requires investing real time and energy in someone else’s development, and that is not how you excel in our business. Or even survive.