Nice to know what people think... but before we all have to read 1001 opinions, leave some space to aswer the question: "I wonder if this is backed by any hard evidence?"

Does anybody know a study that actually compared the success of stats-courses using either SPSS or R. I suppose that the aim and the structure of the course will surely have a major impact on the result, but still, knowing any real study would be a valuable start.

@Andre: "community activity" and publication frequency is not neccesarily related to the (however) measurable success as a tool in teaching statistics. A huge forum and tutorials and mailing lists for getting help can equally well indicate greater difficulties. Also, even the amount of difficulties in using a software is no indication for the "competence in statistics" that should be taught in the courses. I mean on the one hand it may be simple to run a t-test in SPSS on a completely inappropriate problem and on the other hand students may fail to apply a the correct analysis using R. Such proposed indirect measures can be very misleading, especially if they are interpreted unreflectedly or intuitively.

R is one of the best softwares avaliables for any kind of statistics. It gives you the freedom to customize everything, instead of a black box as the closed softwares. There is no limit in what you can do with it. It's definitty worth the learning curve. You will use it for far more than just statistics.

The raw R is not very attractive and hard to use. There are several shells avaliable to improve it´s use, and add some user interface. One of the best avaliable is Rstudio. It will help you to get started.

Andre: the question was: "I wonder if this is backed by any hard evidence?"

Hi,

The R commander is very suitable for  social science students. It  has menu interface ( writing with Tcl/Tk ). There are also some Plugins. One of them have plenty of demos for teaching statistics (second link). I think it can motivate students to use R.

http://socserv.socsci.mcmaster.ca/jfox/Misc/Rcmdr/index.html

http://cran.r-project.org/web/packages/RcmdrPlugin.TeachingDemos/index.html

It´s baked by the huge community avaliable, thousands of books on specific analyses using R, other thousands of books teaching R for different areas and tutorials, videos and classes, foruns and etc. I doubt any software can provide such support for every learning level.

I think SAS has ... .

I think that R is most attractive, and there is a huge community behind it.

There are a lot of plugin, and it could be most useful for searching an employment.

Nice to know what people think... but before we all have to read 1001 opinions, leave some space to aswer the question: "I wonder if this is backed by any hard evidence?"

Does anybody know a study that actually compared the success of stats-courses using either SPSS or R. I suppose that the aim and the structure of the course will surely have a major impact on the result, but still, knowing any real study would be a valuable start.

@Andre: "community activity" and publication frequency is not neccesarily related to the (however) measurable success as a tool in teaching statistics. A huge forum and tutorials and mailing lists for getting help can equally well indicate greater difficulties. Also, even the amount of difficulties in using a software is no indication for the "competence in statistics" that should be taught in the courses. I mean on the one hand it may be simple to run a t-test in SPSS on a completely inappropriate problem and on the other hand students may fail to apply a the correct analysis using R. Such proposed indirect measures can be very misleading, especially if they are interpreted unreflectedly or intuitively.

As Jochen said ability to use the software does not translate to competence in statistics.

The original question included.

"The outcome I would be most interested in is are they able to apply the methods on their own problems?"

I would modify this by adding 'properly' in front of apply. Learning a cookbook may teach you to cook. Learning a statistics cookbook you might learn the methods. Unlike cooking you cannot buy the ingredients at the grocery store.

I've never heard a study making that kind of comparison. But from my experience using a software instead of a high level programming language is choice usually made mainly due to academic tradition.

I'm guessing a bit but I would say that most teachers avoiding using a programming language due to lack of sufficient knowledge in programming. It's true that programming takes a bit more time to learn (but nowadays not that much) but it allows much more control and with that control creativity is stimulated (plot customization, data managing, data processing, etc.).

That said I find that teaching basic programming with a high level language (high level  here is important, meaning no compilation and very little computer related problems) brings far more advantages than using a software.

A small tutorial for R (as an example) made by  a professor for it's class would be enough to easily allow students to do basic operations. Adding small but plenty exercises would lead to programming logic comprehension.

I've noticed that the "introduction to programming" courses in the faculty where I work use this method often. Usually an easy exercise can be a fun exercise and if you have many exercises, odds say you'll eventually make programming errors. Dealing with those brings enhanced understanding. With that understanding it's a small step until students start inventing their own plots and tools. Those "inventions" may even be wrong or inadequate but I would much rather see them trying (and learning from their mistakes) than just following the "always right" software recipe.

Good study material is fundamental from my  point of view. And I congratulate you for even considering "upgrading" your course to teach a tool that will certainly be of great importance in the future of most students.

Sorry for the long discourse. You've touched a really important point as I see it. I find that most faculties are still in the process of catching up with both the modern technical and educational aspects for future professionals (at least in the scientifical areas).

SPSS and other 'panel applications' are for a first study. The code in SPSS is closed (a commercial program) and the windows philosophy with tick that option, but be careful not to tick the same time another option, conflicting with the previous, is a limited practise. Besides that, you cannot share your findings so easy, as with R packages.

Other issues:

1)the fact that you have to write a line command is not so awful as it seems at the first time

2)the windows-habit can be overcome using R-Studio

3)R has better numerical accuracy than SPSS and this means a lot for those who know about computing pseudo-inverse matrices in regression techniques.

Dear all, 

thanks a lot for your thoughts on this issue and they highlighted some relevant aspects for the implementation.

As Jochen commented, my original intention was really to get some references to studies that have tried to test some of these believes, e.g. are students who learned R more creative than those who learned SPSS? What makes a menu interface suitable for social science students? 

All these can easily be turned into empirically testable hypothesis. Maybe this study still needs to be done.

Best

Gerrit

If free software is available for a certain purpose and if this software is appropriate and accepted in the field of research, I do not understand why public institutions do teach and rely on (and sometimes even demand) non-free solutions. (The same fact also applies for operating systems, e.g. Linux vs Windows/Mac, literature management packages, e.g. Zotero/Menedeley vs Endnote/Refworks, and also office solutions, e.g. Openoffice/Libreoffice vs Microsoft Office)

Of course, R is more than just a simple software, as one needs to learn some basic programming. But on the other side, it can be extremely powerful and it does not demand any special (new) hardware or operating system, which is also advantageous for students, as they can install it on their own computer.

I believe a basic knowledge of programming can be very helpful. It gives a new perspective how data can be handled and what is actually behind computer software. If one studies a subject where one is not obliged to learn some programming, learning R is a good way to get at least in contact with programming and at some students further interest could be encouraged.

However, I also don't know of any study you are asking for. But it would be definitely interesting, to run an experiment: Just split a class in half (randomly), set up a clear curriculum (certain statistics and tests) that can be taught equivalently in both software packages and evaluate after the course with a test and some additional questionnaire.

And then, let the students analyze these data...

Dear Gerrit, 

your question is an interesting one and indeed some empirical input on whether R or STATA would be better for teaching than SPSS might help us deliver our lectures better. In my experience as a social sciences lecturer, though, social sciences students in Britain are often confronted with several layers of difficulty when doing statistics. The first layer, and most important, is statistics. The second layer is the dataset and the third layer is the software. When combined, undergraduate (and some postgraduate) students find it very difficult. Hence, SPSS is often used because it does not require students to learn also a new language, but through a friendly interface and with some understanding of the methods they can easily perform their analyses. I am convinced most students have the potential to learn programming and I would strongly encourage it but it seems to be a pragmatic choice and it would much depend on the teaching. Perhaps the answer lies in finding research evidence (perhaps from psychology) on whether individuals find it difficult to program rather than differences between softwares? A quick google search brought up this. Perhaps you can start from here: http://www.psy.gla.ac.uk/~steve/localed/jenkins.html

If difficulty is overcome by good teaching then you may have an answer.

I have taught basic SPSS in the past to a wide variety of students, including Psych majors, and while I have not taught R I do know that a lot of people struggle with it. After more than 12 years teaching programming (but not Intro to Programming) to non-majors I do believe there are some people that just aren't designed to do it, or at least not without a LOT of retraining them how to think differently, and a lot of hand holding.

It may also be affected by whether the course is an "Arts" course or a "Science" course. At our institution they focus on the science of Psychology and it is in the Faculty of Science, not an Arts or Humanities course. This decision drives which kinds of students take the course. Students that are not used to thinking algorithmically, or in highly logical ways will have trouble with programming in R, but will probably be ok using SPSS (or the cheaper, open source version PSPP).

Even then, you will still have people that could perform analysis in SPSS, but not analyse it correctly. You will always have that problem as different students have different skill sets, regardless of the tool you use.

Without any hard evidence to back it up, I will say that students will be able to learn stats better using software such as SPSS because the software is easy to use, allowing a student to focus on the statistics and their analysis. When using R, most students will be struggling with learning the programming, and using the commands, and dealing with computer problems etc., and just trying to get the assignments done, but not learning statistics or getting that knowledge into their brains in a deeper area. 

http://www.gnu.org/software/pspp/

Never heard of such a study. Why don't you collaborate with someone from the Psychology department and do the study yourself? 

It can be a multinational study. For example may be Asian student prefer coding instead of menu driven interface and/or European student prefer visual facilities more than the other nations. The result of this study may be very helpful for both academics and statistical software industry. And I would like to express my interest in joining the team for such a study.

SPSS will definitely be the best option.

@Mohammad: Yes, and it should also consider possible differences between the genders and the main study subject.

Hi guys,

I love R because it is open source, and I teach it using R-studio, which has an interface similar to Matlab and makes it easy for students.

BEST

I think that for applications of statistics, the two softwares fit very well.

The question should not be, which option is easier. The question should be, which is worth the effort to learn it and the price to buy it, for a longer time into the future. And the answer is R!

From "Fox, John, and Robert Andersen. "Using the R statistical computing environment to teach social statistics courses." Department of Sociology, McMaster University (2005).":

"Unlike statistical packages such as SPSS and Minitab, which provide point-and-click graphical-user interfaces (“GUIs”) to most of their statistical capabilities, R is command-oriented."

"It is easier in a command-driven system to correct, modify, and replicate analyses. It is true that many statistical programs (e.g., SPSS) have GUIs that are capable of generating commands, but in our experience users who rely on a GUI rarely take advantage of this feature, preferring instead to point and click their way through the GUI rather than editing commands."

"Those who migrate to R from point-and-click statistical software may not be immediately comfortable with the command-oriented S programming language, but in many respects R is initially much simpler for someone entirely new to statistical computing. More importantly, R implicitly encourages a deeper understanding of the statistical procedures being used than do some other programs, such as SPSS and Minitab."

Best,

Tamara

There are at least three things ("evidence") that make the start easier with SPSS than with R: First, it looks like Excel with "some" bonus buttons. Second, you see the data by default. Third, there is a menu and you do not have to learn commands.

I do a lot of programming on my own, and I switched from SPSS to R two years ago. I never regretted this choice - but it was a hard time. And it is much easier, if (a) you have a basic idea about programming, and (b) about statistics. And the latter is the worst barrier in getting started with R, in my opinion: A newby does not have an idea about data sets, variables, arrays, etc. - "seeing" is a bless then. And a menu is great, if you do not yet know your statistical methods. As soon as you have an ideas about what you're working with and what you're doing, R is really great.

Regarding your question about a direct comparison: You cannot compare apples to bananas. The differences (even regarding getting started) are multi-dimensional :)