I use R for almost everything. The most significant barrier was the initially steep learning curve when I started using it 8 years ago.

To get us started: I use SPSS. It is priced quite high to obtain a personal copy. I would like to use Epistat more. I haven't had the time necessary to devote to learning it as well as I know SPSS.

I use R for almost everything. The most significant barrier was the initially steep learning curve when I started using it 8 years ago.

1)I am using R, although it was very difficult to learn it. (There exist some problems in reading the date formats.)

2)Maple for symbolical statistical computations.(Not suitable for large data handling).

3)I had used SPSS many years ago, but every year you have to buy its new version (Too expensive and too advertised for its abilities.)

4)I have also used Excel, but now I have replaced it with LibreOffice Calc (MS Office was about 150 euro before 3 years, while now a laptop costs 299 euro! It is a clear steal...for such a limited appilcations software)

5)I have a license for SAS, but I don't want to use it, since later I must buy it (Also expensive and yearly paid...)

I use SPSS most frequently. For survival analysis it is fine. However for some specific analyses (like random survival forests) I have to turn to R, because these functions are not (yet) implemented in R.

I use SPSS and IBM SPSS, a problem that I very often faced and solved other way around is that in fitting curve one can fit one dependent variable with an indepent variable. However, if one want to fit a curve (known to fit the given problem) with one dependent and multiple independent variables, there is no option. If there is such an option in SPSS, please let me know.

I use R for most things now, and agree that the learning curve is steep. I probably use it less for statistics than for data manipulation and visualization. The learning curve is steep because most of the documentation seems like it is written for programmers, not for the layperson.

Thank you all for your responses so far. Long ago, I kind of "fell into" SPSS when it was originally conceived and marketed, then for the IBM mainframes. In the meanwhile I've kept up with it... I'm unsure how I would switch to another system given the learning curves. (So interesting that I can program SPSS faster than I can find my way through the pull-downs when so many folks seem to love them.) Regarding other software systems, It's been more a matter of picking up bits & snards as they are used by others around me. Cost isn't such a problem for many of them, they are readily available in my environment. I've enjoyed the growing online availability of specific programs though -- the learning curves are short and cost is null. For those of you who aren't aware of it, visit http://www.statpages.org/ - lots of ideas here.

I use SPSS, Minitab, and Excel (plus Analysis Toolpak). The only barrier i experience in their use is that they do not have commands or features that distinguish non-probability samples and probability samples prior to inferential analysis (tests of significance of difference, relationship). This gives the user the assumption that any kind of sample is fit for inferential analysis.

I have forgotten to mention also STATA, which is very good for econometric analysis and has a living contributing society with many packages and applications.

I have only used SPSS when required to teach or tutor undergrads in research methods or intro stats classes. I hate it, mostly because it seems designed to be the easiest way to employ powerful analyses without having more than the faintest understanding of them. R not only forces the user to do more than button-pushing statistics, but lacks any actual buttons. However, it can be annoying to have to write custom scripts all the time or to be limited to GUIs like RExcel which, while they offer much simplification over using e.g., notepad or the standard R console, are nothing like the e.g., the MATLAB environment which integrates a GUI based around command line scripting with its fully functional and powerful language. SAS I've used on occasion and I have found it ok but I'm neither as familiar with it as I am the above-mentioned 3 nor with others and have not found it offers anything MATLAB & R do not while they do offer significant advantages over it. I'm not a fan of STATISTICA or Stata, but I have found some free statistical/CAS software packages that, at least in tandem with one another, are not bad at all. I wrote about this here:

https://www.researchgate.net/post/What_software_to_choose_for_the_modeling_of_dynamic_process_on_complex_network?cp=re68_x_p2&ch=reg&loginT=ZGZbNlxp-bwNpdZOx9duZfFNunf6PoniAemz2GVZ2D0%2A&pli=1#view=5333b056d5a3f261398b45a2

I use STATISTCA - application is like expanded a spreadsheet, and students likes it.

I mainly use SAS for data handling and statistical analyses. This is mostly because I've invested time in it and have a bunch of small program snippets that I easily can adjust. Can't really recommend SAS though as it still has the feeling of an old-fashioned mainframe system, is rather costly and demands quite some time to master.

Most of my colleagues use R and their experiences are generally positive. I agree with what others have said that you need to be willing to invest some time in it initially (I haven't yet). Among several good reasons to choose R I think that the most important one is, in addition to it being free, that there is a whole community of helpful experienced R users out there. Starting point: http://www.r-project.org/ or (for commercial use) http://www.revolutionanalytics.com/

I always use SPSS, and usually have no problems. At the same time, I use AMOS for Structural Equation Modeling, and I have some problems in calculating Standardized RMR.

I use Stata and I don't see any significant barriers to use it. Sometimes I also use SPSS and Statistica (Statistica has a very good graph editor).

Alan, you are definitely not the first person to ask a question like that. See for instance http://stats.stackexchange.com/questions/33780/r-vs-sas-why-is-sas-preferred-by-private-companies.

I started using Statistica and SPSS about 20 years ago for my first Master Thesis (which places it around 1995). I started using Stata on a new (then) job in 1998, and learned to program it quite well (as you can see from my profile) within may be half a year (although my skillset as a Stata programmer has been continuously expanding ever since). I have been using R and SAS intermittently, but most of the time I just grind my teeth in frustration (especially with SAS), convert the files to Stata and do whatever I need to do. I have been telling myself to learn R, and have sincerely tried on at least five occasions, but the learning curve was too steep every time, and I could deliver the results so much faster with Stata.

One immense advantage of Stata is that you can develop a sophisticated and robust workflow using metadata associated with the data itself, due to the macro language that can be easily picked up by analysts/data users. Workflow means a set of procedures that ensure data integrity and reproducibility of your work, and includes versioning of your data, extensive use of metadata, application of semi-custom flexible programs that can process inputs from a defined class of variables (e.g., detect whether a variable is coded like 1, 2, 3 and then 99, and assume that 99 is a missing value that needs to be converted into a system missing); all of this wrapped up and protected by a revision control system. A really unique feature of Stata is its "assert" command that stops execution when a check fails; this is surprisingly more difficult to achieve in almost any other package, but is crucial for statistical program certification and testing (http://www.stata-journal.com/article.html?article=pr0001). A top notch book on workflows has been published some five years ago (see http://www.stata.com/bookstore/workflow-data-analysis-stata/), but I am not aware of good equivalents for SAS or SPSS. A somewhat comparable approach of reproducible research in R is usually achieved with R Sweave (which is actually much stronger than what Stata has in terms of interacting with text files), but it requires knowledge of LaTeX, which is a long shot in some disciplines/areas. The packages that don't have enough hooks for the end users to write their own code and put bits and pieces of numeric output into LaTeX, Word or Excel files, like SPSS and SAS, are very difficult to deal with once you have spoiled yourself with what Stata or R can do. It's true that you can redirect output from SAS to an rtf file, but these files require humongous amount of editing in the end, as you have to live with what SAS developers had thought was useful for the users to see, rather than what you really need in your project.

I have to add that R is very slow when running loops many times, so for big data simulation studies you have to merge it with a fast language like FORTRAN or C. I don't know why this feature has not been resolved, although its great popularity, so many years now. Maybe its is due to S plus architecture of it.

Thank you everyone - so as I suspected the software we use is a confluence of factors including "matching" our own cognitive abilities to the demands of the software learning curve (preference goes in here somewhere) and the statistical problem we are dealing with... availability of the software generally isn't a problem - most types are available at least on an institutional level. There is a tendency to want copies of software for ourselves, because I suspect that all of us at one time or another want to do something away from work, or perhaps to apply what we know to some other task -- such as introducing another (in my case my son) to the software in order to accomplish something new for them.

I forgot to mention (thank you Stanislav) the ability to edit files be they series of computation/other programming statements or even incorporation of routines or subroutines in external documents. I've found SPSS very friendly for these purposes, particularly by setting up macros in Word documents that execute SPSS in the background and produce the result there in the Word document in a "hot" field... I think there was a software that performed a lot of these functions back in the day under the notion of "document modeling" - HotDocs? - as part of an effort at BYU. They've since gone on in other guises and now aren't especially relevant. I diverge - I like very much the ability to extend my use of stat software into other environments, particularly word-processing type tasks. Have you all experienced this - incorporation of statistical routines into documents such that they continue to report current statistics in designated "hot" fields?

Hello friends, I only seem to be reading papers where SPSS was used. Please let me know of others that I haven't come across, where R, PSPP were used. Thanks.

Dear Alan, latex has many packages for doing what you have mentioned.

I have used Statistica before (for about 3 years) and I now use R for everything (data manipulation and exploration, statistics, graphing and even mapping). As many have mentioned before, there is a steep learning curve and the R literature is not the easiest to comprehend for beginners. Nonetheless, the web is full of very useful help pages on all R topics. Once mastered R is a very powerful tool.

To me, the main advantage of R, apart from being freely available, is that so many packages are available for specialized workflows. It's allways very likely that somebody else has had the same problem as you, solved it and put the solution on the web.

No doubt R software is very often used by good statisticians, but the problem is for most biologists. Its commands are very complicated, not easy to remember. I don't understand why this software still kept under shade of statisticians only. This drawback which lead most biology scholars to move toward SPSS or other related softwares.

R software is my favourite. Especially due to it's flexibility interms of installing your favourite package. Like if you are a finance person, you just download the finance packages.