Will you or won’t you use R in statistical analysis and teaching?

Home Forums Methodspace discussion Will you or won’t you use R in statistical analysis and teaching?

Viewing 15 posts - 1 through 15 (of 34 total)
  • Author
    Posts
  • #4725

    I am interested in how many academic faculty/teachers and researchers out there are now seriously considering using R for their statistics teaching and their own data analysis? Has the recent emergence of a graphical user interface in the shape of R Commander made you more likely to use or adopt R, or are you still happy using SPSS, SAS or STATA?

     

    I look forward to hearing people’s thoughts.

    Patrick

    #4758

    Dear Professor, I am keen to use R for statistical analysis as i have learned from Aligarh Muslim University. Prof.Khan is an expert and provided me all the material needed to start with R free. If you need you can contact him.
    I am planning to use it for my students for marketing research.
    yours sincerely,
    KPrabhakar

    #4757

    Thanks, Dr Prabhakar. I will be interested to hear if others too are making the switch to R and also interested to hear how your students get on with the package during the course. Best of luck.
    Patrick

    #4756
    David Morgan
    Participant

    I don’t teach quantitative methods any more (but I did a grad. level intro to Regression for several years), so here are a some thoughts.

    Even though the final two letter in SPSS are abbreviations “Social Sciences”, I suspect it will become less and less useful for academic purposes. That package has become increasingly business oriented over the last decade, and its recent purchase by IBM suggest that direction is almost certain to continue.

    Can’t say where that leaves SAS, but trying to teach even graduate students how to use “Proc Data” was enough to drive me into the arms of SPSS.

    One advantage of R is that being “open source” allows for a lot of customization. I’m not at all sure whether that includes downsizing it to meet the more limited needs of introductory social science courses (or “right sizing” it for the functions we would need in a more advanced course — which might not include such R functions as extensive modeling options for auto-correlation in time series data).

    But if that kind of customization were possible, the resulting packages could be traded among likeminded colleagues. The same would be true for self-produced manuals — and I could imagine harnessing the kind of collaborative effort that goes into open-source programming itself as a way to produce a modular text as. (I’m thinking of something the old Norusis texts that used to accompany SPSS.)

    Unfortunately for Sage, it would be very easy to distribute such a text for free with pdfs — and combining that with the zero cost of R would be a huge benefit for our students. Either that, or would have to jump very quickly into developing that kind of text with a set of big-name experts, in partnership technical writers who actually understand the needs of classroom professors and their students. And the market would b e even better if appropriate packages of R functionality to match the modules.

    Finally, many folks may not be aware that the larger R project includes something called “Comprehensive R Archive Network,” which contains an incredibly wide range of open-source packages for all kinds of analysis (e.g., for anyone who wants to do Bayesian Hierarchical Clustering).

    ==>David Morgan

    #4755

    Thanks, David. I had not heard of the R Archive Network. That sounds like something I need to look at right now. Its interesting that you feel that SPSS is becoming less of a product for social scientists and more business oriented. I wonder whether others feel that too?
    Thanks again.
    Patrick

    #4754

    Hi Patrick,

    Already use R (with R Commander) for 50%+ of statistical analysis – including mclust for Bayesian clustering methods! – and will be using it as the sole teaching tool for an MSc this September, which I’m sure will be interesting.

    Having said this, it still crashes on me fairly regularly and feedback from the majority of students is that even with R Commander running the interface is too daunting when SPSS/PASW are available alternatives.

    Does anyone use RExcel for high-level inferential analysis?

    –Mark

    #4753

    We have PASW version 18 installed in the Unit and I think the apparent audience changes are entirely a cosmetic branding exercise; I can’t detect any meaningful differences from SPSS v16 (or, really, for that matter, v12).

    There are a number of excellent textbooks available for R (my favourite being Brian Everitt’s ‘Multivariate Data Analysis using R and S Plus’) but I’ve no idea how well they tend to sell; as David says there is a great deal of very good stuff out there for free. Perhaps what’s missing is a mid-level text that bridges programming techniques with commonly-used SPSS techniques (logistic regression, for example, is atrociously difficult to do with any complexity when you exhaust the RCmdr menus).

    #4752

    I now use R for my simulation studies. I am interested in the development of OpenMx in R, which may be a viable alternative to commercial SEM software packages in the future.

    However, as long as my colleagues and collaborators in psychological researches use only SPSS, I can’t switch to R in those researches. If I use R myself, I have to do all the stuff myself. No RAs and student helpers in my team use R, and it is very difficult to hire one who know R *and* psychology.

    For teaching, I am thinking how to include R along with SPSS. My undergraduate students, mainly psychology majors, are expected to know SPSS. Therefore, it is not beneficial to them if I teach R instead of SPSS. However, I believe that if no teachers teach R, then R will never become a popular alternative. Not many students are motivated enough to learn one more tool.

    Any idea on how to introduce R to undergraduates in social science? Or … should R be introduced to them? Too early?

    #4751

    In my MSc. Research Methodology class, The only software we were encouraged and got lessons to use was R. At first, it was difficult for me, but with time and given the internet access, I really dont feel like using any other software. It was a challenge learning R but I consider it a big achievement now. However,had I been introduced to Rcomander before I started off with R, it would have been much more fun.

    Now, I believe R is the way to go.

    #4750
    Phil Murphy
    Member

    Thanks for opening up this line of discussion Patrick. This is a topic that is dear to my heart lately.

     

    We have been using R (R Commander) at the Monterey Institute for two years now. It started as a “hey wouldn’t it be cool if…” discussion and quickly became apparent that it was very possible to have much of the ease of use GUI functionality of SPSS/PASW and the scalability and cost-free (in terms of cash) advantages of R. 

     

    There were a number of reasons for our switch:

    Many of our students are preparing to work with NGOs and other less-well-funded venues. We feel that training someone in this area on something as expensive to use (outside of student packages) as SPSS or SAS would not be responsible. It would essentially put what they had learned beyond their reach once they graduate. Though we do recognize that this is less of an issue
    for those going into large organizations.

    Also, we quickly discovered that R and R Commander are platform independent. Although there can be some annoying bugs in the install process, we have all our students running the software regardless of whether they are on Macs of PCs. (In the past, students would purchase a book without the bundled software or Macs would have trouble with the available bundle.)

    Within the space of a few days of experimentation, we discovered that there is little – if anything – in an introductory statistics course that R Commander cannot do. You can examine and edit your data in spreadsheet format through R Commander. It is just not as pretty and doesn’t have many native functions. Though you can manipulate and transform your variables however you like through the GUI options menu and if you want to manipulate it “spreadsheet-style” then you could always export it as a .csv and play around with it in Excel, Numbers, OpenOffice, Google Spreadsheets, or something similar before importing it again. Additionally, if there is an analytic or graphing function that you would like to do that is not offered in the GUI, there may be an add-on package available and there is always the option of a short scripting lesson.

    This brings me to the scripting advantage of R Commander. All statistical packages offer this, though you will have to turn the function on in GUIs like SPSS. But R Commander makes the scripting an integral part of its presentation and output. With attention, this can be a much less painful gateway into scripting in the R language than starting cold and it will offer those interested in more advanced analyses a distinct advantage.

    Finally, we also offer advanced analysis courses and a social network analysis course. R Commander can do much of the advanced data analytic work and familiarity with scripts will give students a leg up on running some of the more advanced network analytic programs (some of which are developing their own R GUIs) by introducing them to scripting and aiding in the importing of files into R.

     

    I was initially trained in SPSS and functioned fine. Though, since using R and R Commander, I have found myself opting for scripting my personal analyses in R rather than opening SPSS (something that I have not been tempted to do in more than a year now).

    I hope that someone finds this helpful. I should also mention that there really is a lot more help and support for getting started in this than you may suspect. In particular, there is a group called “Revolution R” lead by the creator of SPSS (retired) who offer free support for academic purposes.

    This can be a very rewarding experience for those who are even mildly adventurous.

     

    Good luck in your own explorations!

     

    Phil Murphy

    #4749

    You must know there is a project related with R “Revolution Analytics” is a alternative project using R for business. The leadership of the project is under responsability of Norman Nie, a co-founder of SPSS. Weird, right? Well, some of the members of staff are old workers and executives of SPSS. So, obviously they are not a bunch of stupid guys, they know the future is R and, given they can see the opportunity and Gentleman (one of the R’s creator) is a very ambicious person, all of them are working hard to use R in new commercial ventures.

     

    I think this is very bad for R open source project. But that’s the way it is. Those guys want money and, even thought I think R is one the best softwares I have used, the future of “classic R” is a big uncertainty.

     

    Anyway I use it and I teach it to undergrade and master degree students, sometimes I apply R Commander package, so is quite easy for them work with it (even though R has a very steep learning curve, without doubt).

    #4748
    Jeremy Miles
    Participant

    I don’t see why that’s bad for R.  It doesn’t stop regular R from existing, and the people at Revolution Analytics have produced add-ons to regular R, which they have released as open source.

    They are solving one of the problems that R has which stops it being adopted, which is that there is no support available.  A large corporation wants support before it adopts a program, and it’s willing to pay for it. 

     

    There have been several releases of ‘Classic R’ since Revolution R has existed.  

    #4747
    Jeremy Miles
    Participant

    ?? Why?

    #4746
    Jeremy Miles
    Participant

    One thing I use R for is SEM.  SEM add-ons to programs are either expensive (AMOS, for SPSS), not very good (CFA1 for Stata) or complex (tcalis, for SAS).   The sem add-on for R is free, easy to use, and easy to teach.

    #4745

    Hi,

    I guess i should not be posting this here. If its misplaced, I am sorry. I am a newbie when it comes to R . I have used SPSS and do understand basics of statistics, though not a mathematically minded person by any standards. Nor do I have experience programming computers though I am pretty comfortable with fairly advanced level stuff. Now what I need is a way to learn R from the scratch …….preferably online. The immediate problem I have on hand is test development (psychometry) hence item analysis and multivariate analysis are things I will need in the medium term. I am willing to take time to learn. Would REALLY appreciate it if there is anyone who is willing to help me start and answer some email questions (maybe stupid ones) as I go along.

Viewing 15 posts - 1 through 15 (of 34 total)
  • You must be logged in to reply to this topic.