1
$\begingroup$

A doctor friend has asked if I would teach them some basic programming, because they think it will be useful when they start to do research in the future.

To give some background - they don't know what language their lab-mates will use (because they don't know what lab they are joining) but they would like to learn something that will be generally useful. They are not interested in doing statistical research, but rather applying statistical techniques to medical data.

There are some obvious options (I know enough about all of these to teach them)

  • Python
  • R
  • MATLAB

and a few less obvious ones (I don't know anything about these)

  • Stata
  • SAS
  • SPSS

I'd like to know what languages are most commonly used in medical statistics - do people use general purpose languages like Python, or more specialized languages like R and MATLAB, or statistical software like Stata/SAS/SPSS?

I should clarify that I'm thinking of someone whose primary job is as a doctor/medical researcher who will be applying statistics, not someone who is a professional statistician or software developer who happens to be working in medicine (so e.g. I think that C/C++/Java would all be poor choices).

$\endgroup$
  • $\begingroup$ I think you are missing SAS in your list. And I would hesitate to teach a language, I know only a bit about. $\endgroup$ – Roland Oct 31 '13 at 10:14
  • $\begingroup$ Edited to include SAS. This question is just to get an idea of what languages are actually used. Obviously I wouldn't attempt to teach one I hadn't used before (also, when I say I know "a bit" about Python/R/Matlab I am being modest... I use them all regularly in a production environment). $\endgroup$ – Chris Taylor Oct 31 '13 at 10:17
  • $\begingroup$ This could do with some sharpening up: it seems to blur "medical research" and "medical statistics" (or "biostatistics"). Perhaps most fundamental medical research is not statistical, or at least not strongly statistical. $\endgroup$ – Nick Cox Oct 31 '13 at 10:26
  • $\begingroup$ I've never heard of medical researchers using EViews. Its loyal fans seem mostly to be economists. I don't think it claims to be general statistical software. $\endgroup$ – Nick Cox Oct 31 '13 at 10:32
  • $\begingroup$ @NickCox thanks for the comments, I'll see what I can do. $\endgroup$ – Chris Taylor Oct 31 '13 at 10:35
2
$\begingroup$

This question seems to me to conflate two issues: Programming and statistics.

I don't know what programming languages are used in medical labs, although I get the sense that none are. In terms of statistics, I'd say R and SAS dominate. These are radically different languages. The problem with trying to teach SAS is 1) Since it isn't on your list you probably don't know it and 2) You'd have to have access to it, and it's expensive. That would lead to teaching R. But if they wind up in a place that uses SAS exclusively, I don't think R would help much.

However, if the person will be doing their own analysis, then R is fine. I would try to teach reproducible research methods.

$\endgroup$
  • $\begingroup$ I don't know that anyone has hard data on usage, but Stata is very important to many groups in medical statistics or biostatistics. $\endgroup$ – Nick Cox Oct 31 '13 at 10:29
  • 3
    $\begingroup$ I agree with you Peter, but I am slightly more optimistic. An important aspect of using R/SAS/Stata/SPSS/... is the statistical way of thinking. Learning how to use such a package will require the student to learn both the specific package but also the general statistical way of thinking. I'd say that learning to ask the right statistical question is more useful (and harder) than learning the language to actually pose the question in. In that regard, learning any statistical package will be helpful even when the student will end up using another one eventually. $\endgroup$ – Marc Claesen Oct 31 '13 at 10:32
  • $\begingroup$ I agree with @MarcClaesen, but think you should be even more fundamental. In my experience people without any programming or mathematical experience often have more problems with learning how to develop an algorithm than with implementing it in the language of their choice. $\endgroup$ – Roland Oct 31 '13 at 10:57
  • 1
    $\begingroup$ Interesting @MarcClaesen . I am not sure I agree with you, but it may depend on what you mean by "learning how to use such a package". I have seen plenty of analyses in SAS and SPSS that exhibit a clear lack of a statistical way of thinking! (My impression is that the analyses I've seen in R tend to exhibit less of this lack, but my sample size is quite small). Any computer package can foster the "The computer said it, I believe it, that settles it" mindset. $\endgroup$ – Peter Flom - Reinstate Monica Oct 31 '13 at 10:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.