9

I am currently a statistics student within a very good program. We use Minitab for a variety of items/classes. But, as my undergrad taught me, what you use in school isn't necessarily what is used in the real world. I'm also well-versed with C#, python and I'm learning SAS (almost ready to take the first certification test for SAS).

Is Minitab used extensively in real world studies/jobs? Should I invest more time in learning SAS instead?

Unknown Coder
  • 166
  • 1
  • 1
  • 7
  • 3
    That academia/real world distinction may be more fuzzy than you think. That said, I can't imagine (and I've never experienced) anyone asking for Minitab for anything. – conjugateprior Sep 15 '14 at 20:21
  • 2
    Of the statistician/analyst/programmer-analyst job openings I've seen, I'd guess 10-15% either require or prefer SAS skills. For Minitab, it's <1%. Then there are those that say something like "experience in a statistical program such as SAS, SPSS, or R" and few of these mention Minitab. – rolando2 Sep 15 '14 at 23:05
  • 2
    The utility of various statistical software strikes me as potentially on-topic (although there can be no doubt it is potentially off-topic as well). Thus far, this thread does not seem to be drawing answers that are primarily opinion-based. – gung - Reinstate Monica Sep 15 '14 at 23:13
  • One thing I did not want to put as an answer but more as a comment : The need to code properly is very important. A lot of statistician I met do not know how to do it e.g. no comments, no space. I know this is not what is required in statistics but it does help for whoever read your code in the long run. – Andy K Sep 16 '14 at 05:50
  • 2
    Minitab can be good for quality control and other industrial statistics. I know some mining companies use it! – kjetil b halvorsen Sep 16 '14 at 11:43

4 Answers4

12

Rexer Analytics does a tool survey every year that you can obtain by e-mailing them. This is not the best data in the world, but Minitab is pretty far down the list, though its users do seem to like it. The rarely-seen-in-the-wild characterization is consistent with the better Muenchen data from job postings and related sources (including the Rexer data), and my own experience in industry research.

Based on the above data, I would spend your time learning R, unless there are some industry-specific reasons to focus on SAS. I write this as a heavy and happy Stata user that has never had an employer balk at purchasing a license. An added complication is that most people around you will use Excel for just about everything and you should learn a tool that plays nice with it, as well as being able to query SQL databases.

dimitriy
  • 31,081
  • 5
  • 63
  • 138
7

Minitab's used a lot in production engineering, quality control, & Six Sigma (& when engineering companies use it in these areas, it may also have become the default statistical software in others).

Based on my experience, I reckon graduates worry too much about software (& not enough about other things, especially consulting skills); demonstrated competence in statistical programming is generally important, but lack of familiarity with any particular language/software needed for some job is easily dealt with after starting it. I would say, though, that a Minitab (or SPSS) user, rather than a SAS (or R, or Stata) user, has perhaps to take pains to show that they can do more than point & click to run canned analyses—e.g. writing macros for non-linear regression, or whatever's not in the menus at the moment.

Should you invest more time in learning SAS instead? Instead of investing more time in learning Minitab?—probably yes. Instead of investing more time in anything that gives you experience of working with real data on real problems &/or collaborating with domain experts in another field?—probably no.

Scortchi - Reinstate Monica
  • 27,560
  • 8
  • 81
  • 248
  • Quite right about the worrying, you have to use what the shop uses. z/OS shops tended to use SAS, VMS shops used Minitab. The whole argument is very similar to the WinZip/WinRAR/gzip debate. Different wrappers around similar publicly available methods. R vs SAS, Haskell vs Python. Unless it is *your* startup, you work with what you are told to. – mckenzm Jun 29 '17 at 06:13
6

Minitab is very popular for reliability and warranty analysis. Used all the time in these areas, especially for litigation purposes. I know a guy who has used Minitab since it was just a command line prompt. What he can do with it in a short time is very impressive.

It's hard to say what to invest more time into unless you have a specific area you're interested in. Minitab is relatively easy to pick up compared to R or SAS.

Glen
  • 6,320
  • 4
  • 37
  • 59
4

Google Scholar search (9/15/2014) for:

     Program      Hits
         SAS 2,610,000
SPSS or PSPP 1,640,000
       Stata 1,280,000
  Statistica   459,000
         JMP   249,000
  R and cran    86,500
     Minitab    85,800
      Systat    73,800
        BMDP    45,900
      SUDAAN    17,100

That said: there's lots of reasons besides popularity to choose a platform:

  • performance for specific data set sizes
  • cost
  • extensibility
  • how fast new techniques are released/old techniques are updated
  • documentation and support
  • perpetual versus rental license
  • what your team uses
  • portability
  • multi-core support
  • multi-user support
  • the nature of the errors in the software or documentation
  • does it do specifically what you need it to do
  • interface
  • &c.
Alexis
  • 26,219
  • 5
  • 78
  • 131
  • 6
    I believe that your result for R is skewed by your search term. Many publications don't mention CRAN, but write something like "R software", "R statistics environment" or similar terms. The search term "R package" gives 99k hits already. – Roland Sep 16 '14 at 09:56
  • @Roland A similar situation applies to the other statistical computing platforms. My motivation is in providing a very rough sense of whether these packages are professionally used (they all are), and a crude idea about the *order* of their frequency of use. You approach does not improve on that for my purposes. – Alexis Sep 16 '14 at 13:58
  • 2
    I was trying to say that I do not really believe the order you found with this approach. The reason being that the other platforms have much better search terms (not your fault that "R" is so ambiguous). – Roland Sep 16 '14 at 14:05
  • @Roland And I meant "order" as in "on the order as" 79k is on the same order as 99k. Neither 79k nor 99k are on the same order as 2.6m. That said: yes: I agree with your concerns about "order" meaning "ranking". – Alexis Sep 16 '14 at 14:31
  • 3
    I would expect R to be somewhere between 3 and 5 in rank (if you include Excel as number 1). I have trouble believing that SAS is in front by such a large margin (or at all). But of course I can't prove it. However, when I was an active Wikipedia contributor I learned that the number of hits reported by Google cannot be trusted (some people were trying to use these numbers to prove "relevance"). – Roland Sep 16 '14 at 14:39
  • @Roland Google Scholar, not Google. – Alexis Sep 16 '14 at 14:50
  • I know, but presumably it's the same algorithm. – Roland Sep 16 '14 at 15:06