Use of R for biostatistics publications

Question

Presumably this has been discussed already, but I wanted to get some sense of how prevalent is the use of R for biostatistics publications. On PubMed I did find a number of research papers that were conducted using R (as mentioned in the Methods or Statistical Analysis sections of the documents). However, the majority were done using SAS/SPSS and also an appreciable number using Stata.

There has been a lot of talk about R in recent years and I am looking for some data that can corroborate the fact that R has been rising in popularity in the commercial area. In my experience, however, I have been seeing some resistance for firms to switch to R from SAS, even though there is so much more that you can do with R compared to SAS. There is also the concern that FDA "prefers results in SAS" ... although I have not seen anything officially mentioned by FDA to this effect.

I understand that SAS has been there for decades, it is what many R&D departments are most comfortable working with, etc., etc., ... and am looking for real-world examples of who are instead using R for professional research and/or publications.

As you say, much discussed e.g. http://stats.stackexchange.com/questions/33780/r-vs-sas-why-is-sas-prefered-by-private-companies See also the site http://r4stats.com/ (which I find a little partisan, but everyone has biases of some kind). — Nick Cox, Jul 19 '13 at 10:40
Unless there is something you must do with one program you cannot do with other, if the research is good any software will be ok (as you mentioned you have seen papers with all of them). But my comment was mainly to ask what does the acronym FDA mean? Thank's. — Andre Silva, Jul 19 '13 at 13:41
+1 to Nick's suggestion and Andre's comment. Speaking as a PhD student, R is by far the most commonly used environment used by students today so just for publication purposes probably that won't be a problem. Major packages also are usually peer-revieed in the Journal of Statistical Software or similar publications. I don't know about professional research. (@Andre: U.S. Food and Drug Administration). — usεr11852, Jul 19 '13 at 13:49
@user11852: Pleased that you agree with me, but I want to qualify your statements. Of thousands of R packages, I suspect that the peer review most received is that many have banged hard on them and some have tried critical checks. For that matter, the peer review received by SAS, SPSS, Stata, etc. is mostly of similar kind. Also: what is your sample of "students today"? Patterns of usage vary much between disciplines. The software I know best (Stata) is used intensely in biostatistics, economics, econometrics, political science, sociology, less by e.g. ecologists, mainstream statisticians. — Nick Cox, Jul 19 '13 at 14:15
@NickCox: Statistics. I should be more precise on that, you are correct. For jstatsoft reviewed packages I think "banged hard and tried critical checks" is a serious understatement. I have seen reviews that had line-by-line comments. I did not mean to imply that R is significantly better in terms of quality but to dismiss the idea of "home-brewed" code (for big packages at least; for example concerning LME models, MCMCglmm, Rcpp , lme4, glmmADMB, are definitely not a lone cowboy-coder's brainchildren). — usεr11852, Jul 19 '13 at 15:03
@user11852 We agree on all fundamentals. My point is what peer review the thousands of R packages not reviewed by JSS have received. — Nick Cox, Jul 19 '13 at 15:20
@NickCox: I definitely agree with you on that. I never said "all" or "most packages" though; I specifically said "major packages". And yes, of course, smaller packages that were submitted a couple of years back and never received any updates probably aren't the safest to go around. :) — usεr11852, Jul 19 '13 at 15:40
I did notice your "major". I don't want to disagree at all, but I wonder what the result would be if you asked experienced R users to name their top 20, 50, 100 R packages and then checked how many had been through JSS. That's not a criticism in any sense. — Nick Cox, Jul 19 '13 at 15:43
Attempting to be neutral here: (1) Look around in the sub-field you work in to see what people use most. That rarely rules out using other software; it just means that you may experience entrenched attitudes against it. (2) Whenever anyone maligns other software a fair question is always "and exactly how much have you used it?". If not at all, you're hearing second-hand prejudice. If not much, that is no basis for well-informed and reasoned comparisons. (3) Being negative about software X only because X is not Y and you know and like Y is not a serious argument. — Nick Cox, Jan 28 '18 at 13:22

score 0 · Answer 1 · edited Jan 28 '18 at 13:15

You ask two very different Qs:

1) "R has been rising in popularity in the commercial area"?

2) "How prevalent is the use of R for biostatistics publications"?

Regarding 1), I'll put it bluntly: despite all this tedious talk about R, I have worked in 5 countries across 3 continents and inside the industry I never once heard anybody refer to R. I worked for very large multinational companies and startups, and CROs and pharma. They use SAS. Simply look at the industry association for prgrammers, i.e. Phuse. And what are they discussing at their conference? And what are they asking for in the job adverts? E.g. pharma jobs The first job I looked at says: "Extensive SAS programming expertise to an advanced level". Says nothing about R, but this does not stop these impassioned and incessant R programmers from claiming the imminent death of SAS.

Regarding 2), academics are much more interested in publishing and they are much more interested in R; thus you are fine to use R in this case, no problem there.

Use of R for biostatistics publications

1 Answers1