Does anyone know of research showing the different types of statistical software used? I'm looking to find out things like what percentage of people use SAS vs. Stata, R, or other packages. I'm be happy to see references not only for statistics users but users in other science fields that make heavy use of statistics.

  • 3
    $\begingroup$ Take a look at the links in my answer here. Also, it is Stata and not STATA. $\endgroup$ – Dimitriy V. Masterov Jan 14 '16 at 0:31
  • $\begingroup$ Yes of course. I think I got a little carried away with capitalization there! Thanks! $\endgroup$ – StatsStudent Jan 14 '16 at 1:37
  • $\begingroup$ You can find an overview of such packages in this blog posted on jWork.ORG (2018) $\endgroup$ – Tanuasha1 Jan 19 at 1:32
  • $\begingroup$ It is very domain specific, e.g. SAS is very popular in finance, SPSS in social science, python for machine learning etc. $\endgroup$ – Tim Jan 19 at 9:20

Number of users is very difficult to determine for free software like R, since there are servers all over the world, so even if one kept track of downloads on one site, the rest would be unknown, and people can share copies around; a single download of R might potentially be used by hundreds of people. One can perhaps get some idea by tracking discussions on various sites or mailing lists, e.g. by looking at volume of questions on Stackoverflow (and various other fora, especially since some programs have dedicated discussion)

Bob Muenchen produces regularly updated measures. At the time of posting the most recent one was October 2015:


There's also other surveys. For example, David Smith posted about two such surveys in Novenmber 2015, here:


  • 4
    $\begingroup$ The measurement problems are everywhere dense in this space. Suppose there is a university licence for some software. Companies could hardly cite anything except the number of students who in principle have access to that software. I download R, which I use about once every 3 months, more frequently (because it changes so much) than I install the software I use daily. All vendors and producers have understandable biases to stress the measures that show their wares in a good light. $\endgroup$ – Nick Cox Jan 14 '16 at 14:59

This site is primarily of concern for users of software for statistical purposes On this site as of today, 2019-01-19, the following package tags have the listed number of followers:

r           19572
Matlab       1158
Python       2253
Mathematica    43
SAS           597
SPSS         1613
Stata        1091
Fortran         5
C++            39
Excel         359
Perl            5
Oracle          8

Any others you can think of, just search the tags for it.

  • $\begingroup$ Very helpful, @Carl. Now why didn't I think of that!? Thank you! $\endgroup$ – StatsStudent Jan 19 at 16:07
  • $\begingroup$ If so, you might just accept the answer. It does not affect reputation because @whuber made this post into a community wiki on Jan 14 '16 at 2:02. $\endgroup$ – Carl Jan 19 at 22:43
  • 3
    $\begingroup$ It's only an anecdote -- but I don't follow any software tags on the grounds that software details are off-topic here any way. So I don't even follow the software I use most. A bigger deal is how far CV is the most likely forum even for statistical questions faced by those using particular software. That may be so for R; I doubt it's true for SAS, SPSS, Stata. or any software where there is software-specific support that also covers statistical questions. $\endgroup$ – Nick Cox Jan 20 at 11:51
  • $\begingroup$ @NickCox All surveys are anecdotes in that there is a difference between a first guess as to meaning of a survey's results and a second guess about its meaning. One can, I suppose, more easily correct for bias than obtain unbiased results, but in this case the value of the post was to point out that CV is a source of information about the question, with its own peculiarities to be sure. $\endgroup$ – Carl Jan 20 at 20:36
  • 1
    $\begingroup$ @NickCox We have that same anecdote in common, so $n\geq2$. $\endgroup$ – Carl Jan 22 at 5:24

The KDNuggets blog does surveys of software used as reported by participating data scientists. Typically, these surveys are annual. Note that the percentages they report are not the same thing as market share of software sales, it's just a proxy for that metric. Moreover, it may or may not reflect the specific languages that interests you.

Here is a link to the results from a recent survey report:


But they have many more trendable polls available. Check on their site.

  • 1
    $\begingroup$ As with all surveys, the question of what are the target population and the sampled population is paramount. I have no evidence -- that is all part of the same big problem -- but my guess is that thousands if not millions of students and researchers using statistical software would never even hear of such a survey. But what is the aim here? If it's to catch every student user of some software using a university site licen{s|c}e who did one or two statistics courses and then moved on, it's probably going to fail. $\endgroup$ – Nick Cox Jan 22 at 9:55
  • 1
    $\begingroup$ None of that is to deny that little indications of what's hot and what's not aren't interesting, even intriguing. Comments on social media about different statistical or related software that I stumble against often seem dominated by memes on what is in, is out, going up, going down, regardless of inherent merits or whether the person making the comment has deep or wide experience on which to base a comparative opinion. $\endgroup$ – Nick Cox Jan 22 at 9:58

You can find an survey of data-analysis and statistical software in this this article. The result of this survey is shown in this image. Note that it is based on various blogs enter image description here

  • $\begingroup$ Hmm. Thanks for posting, this but I consider this very suspect and would urge users caution when examining this graphic. For one, it seems to be a very small sample size, although it's tough to tell since there doesn't seem to be a sample size reported. Nor can I find anywhere where the methodology is explained. This appears a bit more like a marketing ploy, masquerading as potentially useful data. $\endgroup$ – StatsStudent Oct 3 at 0:47

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.