# [R] Is R good for not-professional-statistician, un-mathematical clinical researchers?

Andrew Robinson andrewr at uidaho.edu
Fri Aug 20 01:05:46 CEST 2004

Jake,

this is a great question.

I have experience teaching R to forestry and natural resources graduate and undergraduate students.  These students are amongst the least numerically comfortable students I've taught.  I have found that the numerically comfortable among them leap into the challenge, and the remainder reject it (and usually me) as irrelevant and deliberately obtuse.  I am heavily biased towards R because every once in a while I find myself writing code with a big happy smile.  It's just fun.  There are some things that it does, that it does just beautifully.  Oh, and I'm addicted to Sweave :).

R, by its nature, exposes the seamy underbelly of statistical reasoning, and this creates various degrees of empowerment and fear, depending on the student.  R requires us to create the syntax of our commands, and to think through the process of analysis before its execution, at least enough to compose a coherent statement.  Of course, this is educationally a very good thing, but also very challenging.

I would suggest that it's just as plausible to teach R in a pre-packaged way as it is to teach the other applications in a pre-packaged way.  The problem arises when the students have to do their own analyses.  R has many elements that require confidence and experience to overcome efficiently.

For example, let's take error reporting: one really has to have at least a grasp of matrix algebra to know what singularities are.  Yet, singularities will be reported in errors for tools that, prima facie, have no necessary obvious relation with matrices.  Or another example is: we don't need to know qr-decomposition or svd in order to be able to understand the statistical elements of a multiple regression.  But, sometimes (NOT in those cases, I hasten to add, but they're a good example of the kind of thing I mean) these details bleed through.  So, our error messages sometimes, lack obvious statistical relation to the problem at hand for the neophyte.  They can seem cryptic and obscure.

Ok, this is inevitable, in a community-generated product like R, but it is a hurdle that students will find frustrating, and will take them a long time to overcome, regardless of their good intentions.  Many of the help files include references to further reading, which is excellent and essential, but some do not.

Now, I argue that student will certainly benefit from adopting the hacker-style can-do attitude necessary to plough forwards.  But they rightly ask: is this the most appropriate medium for that approach to be encouraged, for us?  A problem is that for success, they not only require the hacker-style can-do attitude, they also require technical background, which they do not intend to develop.  Googling alone is not the answer, nor is R-help.  So it depends on the student.

In general, though, I don't think that R by itself can be considered adequate for not-professional-statistician, un-mathematical clinical researchers.  I don't think that it wants to be, or that we want it to be, enough.  So, increasingly, I do think that we should learn another language, and offer such students the option of a more unified approach.

Thanks for a very thought-provoking question.

Andrew

----- Original Message -----
From: Jacob Wegelin <jawegelin at ucdavis.edu>
Date: Thursday, August 19, 2004 4:45 pm
Subject: [R] Is R good for not-professional-statistician, un-mathematical clinical researchers?

>
> Alternate title: How can I persuade my students that R is for them?
>
> Alternate title: Can R replace SAS, SPSS or Stata for clinicians?
>
> I am teaching introductory statistics to twelve physicians and two
> veterinarianswho have enrolled in a Mentored Clinical Research
> Training Program.  My course is the
> first in a sequence of three.  We (the instructors of this
> sequence) chose to teach
> R rather than some other computing environment.
>
> My (highly motivated) students have never encountered anything
> like R.  One frankly  asked:
>
> "Do you feel (honestly) that a group of physicians (with two vets)
> clinicians will
> be able to effectively use and actually understand R? If so, I
> will happily call this
> bookstore and order this book [Venables and Ripley] tomorrow."
>
> I am heavily biased toward R/S because I have used it since the
> first applied statistics
> course I took.  But I would love to give these students some kind
> of objective information
> about the usability of R by non-statisticians--not just my own bias.
>
> Could anyone suggest any such information?  Or does anyone on this
> list use R who is
> a clinician and not really mathematically savvy?  For instance,
> someone who doesn't
> remember any math beyond algebra and doesn't think in terms of P(A|B)?
>
> Or have we done a disservice to our students by choosing to make them
> learn R, rather than making ourselves learn SAS, Stata or SPSS?
>
> Thank you for any ideas
>
> Jake Wegelin
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help