Re: st: Re: Your paper on Stata,SAS and SPSS

[Alan Acock <acock@mac.com>]

John Hall indicates that with the limited math/statistics background of his students, the "syntax examples I have seen in Stata, . . . would easily put (them) off." I'm not sure what examples John has seen. Often those shared on statalist are technical and not appropriate to beginning students. If you think about a basic introductory research courses in the social sciences, the actual Stata commands that would be relevant are vastly simpler than those used in SPSS or SAS. For SPSS readers who don't appreciate this, here are a few
examples they might compare to the lengthy syntax needed by SPSS

(chi-square test)
.table var1 var2, chi2

(independent t-test)
.ttest var1 var2

(grouped t-test)
.ttest var1, by(gender)

(Correlation matrix)
.corr var1 - var10

(OLS regression)
.regress y x1 x2 x3, beta

(logistic regression)
.logit y x1 x2 x3, or

If they get a little more advanced and do a Poisson regression, compare the SPSS command to do a Poisson regression to Stata's--Really--try it in SPSS
.poisson y x1 x2 x3

How about doing a principal component "factor" analysis--the default type in SPSS. In Stata the command is:
.factor x1 -x10, pcf
Then, if you want a varimax rotation you have the post estimation command
.rotate

Stata's goal is "type a little, get a little." The output wastes far less space than does SPSS. There are options and post estimation commands for more advance users. For example, if you want alpha for a 10-item sale you would enter 
.alpha var1 - var10
If you wanted item analysis like SPSS or SAS provide, you would enter
.alpha var1 - var10, item

Of course, Stata has excellent menus, but the reason Stata users don't use them as much as SPSS users is less the acknowledged elegance of the SPSS menus than it is the rigid structure of Stata commands resulting in much simpler command structure (syntax) than is found in SPSS. I suspect that if John listed all the procedures he taught in his research methods course, the Stata commands could be all listed on a single page.

Although Stata is vastly more powerful than SPSS for advanced statistical applications (as well as a fraction of the cost), Stata is also much less cumbersome for elementary applications. I was motivated to write my "Gentle Introduction to Stata" (StataPress), because many books about Stata focused on advanced applications and many social/behavioral scientists had the notion that Stata was just for advanced users. I believe that the advantages of Stata are even greater for beginners--not to mention offering them greater opportunity for growth.

Alan Acock


On Aug 10, 2010, at Tue Aug 5 2:02 , John F Hall wrote:

> Michael
> 
> Thanks for your prompt and appreciative reply.
> 
> For the kind of students I used to teach (no previous computing, statistics (or even much maths) I'm still not convinced about Stata.  Most of them came from backgrounds in sociology and related subjects and Stata seems to me to be heavily statistical.  From the syntax examples I have seen in Stata, they would be easily put off.  However, modern students are very different: all have their own computers or laptops and I'm targeting the ones with PCs, Windows and Word (those with Mac and Linux will have to wait for someone to convert my tutorials, but there's nothing to stop them having a quick peep).
> 
> I can see why some people in "survey research methods" are switching, but they have a very narrow definition and they're more into the statistical aspects such as sampling bias, non-response etc, rather than the substance of the survey.  In the UK that has always been a major difference in the definition of "survey methods".  It's unfortunate (a bit like the false distinction between qualitative and quantitative) but I'm afraid technique currently has dominance over content in some quarters.
> 
> There has also been an interesting exchange between a bunch of Brits on the relationship of the then Social Science Research Council (funding agency) to the development of quantitative methods, with special reference to sociology.  Jennifer Platt (Emeritus Prof of Sociology, Sussex) is the official historian of the Briitish Sociological Association and recently presented a paper at the World Congress of Sociology, Gothenberg, in a session for the Resarch Cttee on the History of Sociology about deliberate attempts to change the direction of sociology.  She made the mistake of contacting me (in my capacity as Editor of Quantitative Sociology Newsletter, which ceased publication in 1984).  I haven't seen her for 20 years, but I have a vivid (if not always 100% accurate) memory and managed to track down most of the people who were active in promoting or enabling quantitative methods and/or computing in sociological research in the UK in the 1970s.  Poor Jennifer is now buried !
 in mounds of fascinating, detailed and learned reminiscences!
> 
> Part of this exchange had some snide references to "plumbers" (computer and statistics people) who could be called in to help out if thought necessary by the superior intellects of "sociologists".  I retorted, "The late Angus Campbell (Director of ISR, Ann Arbor) once remarked to me that you wouldn't expect a chemist to work without knowing how to put a retort stand and tubes together, so why should sociologists not be expected to have at least a few basic technical skills?   At PNL I used to explain my job as teaching sociologists how to count.  At both SSRC and PNL I and my staff upset a lot of people by turning round jobs in 3 or 4 days (sometimes being specially called in) that they had been messing about with (wasting taxpayers' money or funding agency's patience)  for months, if not years, too proud or ignorant to seek advice or assistance: others were eternally grateful, but you can't please everyone."
> 
> There were also some snide comments about research units and centres springing up like mushrooms to cream off research funds, but that's a whole new story.  If anyone's interested, I can forward the relevant selections.
> 
> John Hall
> http://surveyresearch.weebly.com
> 
> 
> PS  I've copied in parts of other mails so that you and others can make sense of your reply.
> 
> ----- Original Message ----- From: Michael N. Mitchell
> To: johnfhall@orange.fr
> Sent: Tuesday, August 10, 2010 4:41 AM
> Subject: Re: Your paper on Stata,SAS and SPSS
> 
> 
> Dear John
> 
>  Thank you so much for your email. My apologies for my delay, I have been buried with
> many things, including focusing my efforts on my book writing.
> 
>  I am delighted that you have been engaging this issue of Stata and SPSS and have been
> fostering some cross communication of the communities via the SPSSX and Statalist
> listservers. Having used both packages for many years and having followed each list for
> quite a while, I know how each community can be very isolated from one another. And, I
> especially understand the issues that the SPSS folks are dealing with and commenting upon.
> That is part of the reason for the technical report that I wrote, trying to help people
> take a wider view of what is available. For those who have access to multiple packages, to
> encourage them to use the best of each tools from each package, and for those who are
> using a single package, to consider the alternatives and to consider whether you might
> want to make the effort to switch to another that might, in the long run, serve you better
> than the package that you know. At my work, I have had multiple people make the switch
> from SPSS to Stata and very quickly they do not look back. And, the cost difference is
> astounding. For the price we pay for one SAS license or the cost of about 2 SPSS licenses
> we get about 30 Stata licenses.
> 
>  I am no longer with UCLA so cannot assist with "web link exchanges", but I am sure that
> the UCLA stat group would be very interested in this. You can write to them at
> ATSstat@ucla.edu .
> 
> Warmest regards,
> 
> Michael N. Mitchell
> Data Management Using Stata      - http://www.stata.com/bookstore/dmus.html
> A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
> Stata tidbit of the week         - http://www.MichaelNormanMitchell.com
> 
> 
> 
> On 2010-08-05 10.31 PM, johnfhall@orange.fr wrote:
>> Michael
>> 
>> to A desperate user in Spain whose university will discontinue SPSS 15 next year and not rplace it.  She started a huge and ferocious debate on the on the SPSSX listserver about IBM/SPSS business models etc.  In a reply someone just posted the link to your paper.
>> 
>> I haven't worked through it yet, but from the thoughts in your abstract and acknowledgments I detect a kindred spirit working in familiar territory.
>> 
>> I have used SPSS on dozens of surveys and thousands of queries since 1972 and am currently working on a stack of learning materials from the postgraduate Survey Analysis Workshop (part-time, evening) I designed and taught from 1976 until I (early) retired in 1992.
>> 
>> These were for various releases of SPSS on a range of machines culminating in SPSS-X 4 on a Vax cluster.  Since 2006 I have been updating and expending these to use with SPSS for Windows on a PC (which involves conversion from WordStar4 to Word and a switch from DOS to Windows, neither of which I had ever used before).
>> 
>> Since September last year I have been developing a new website and have now uploaded a substantial body of entry-level SPSS tutorials, exercises and specimen answers.  They use syntax in preference to the drop-down menus, but many examples are also repeated using the menus.  They are oriented towards survey research rather than statistics and are aimed at teachers, researchers and students with little or no previous experience of statistics (a sort of "Clod's Guide to Survey Analysis Using SPSS"). So far have avoided using a single equation, but that may have to come when I get round to explaining inferential statistics in the later stages of the course.
>> 
>> There are many SPSS courses around, but mine is different, and perhaps more fun.  The website is http://surveyresearch.weebly.com/index.html and as well as SPSS it it also carries a wealth of research reports and other materials.
>> 
>> Many of my colleagues in the UK are now switching to Stata, but I think SPSS is far more suited to the kind of material I'm handling.  However, I'd be interested in seeing parallel Stata syntax and output for some of my examples.
>> 
>> I already have a link to ATS on my SPSS intros and tutorials page http://surveyresearch.weebly.com/spss-intros-and-tutorials.html.  Once I've had a look at your paper, would you be happy for me to add a link to it from my site?
>> 
>> Finally, there's an account of how I got into all this in my Old Dog, Old Tricks presentation (http://surveyresearch.weebly.com/7-old-dog-old-tricks.html) and the first slide show is fun.  There are also two extended critical reviews of Julie Pallant's "SPSS Survival Manual" (2001 and 2005, both different) on http://surveyresearch.weebly.com/8-spss-text-books.html
>> 
>> Regards
>> John Hall
> 
> ----- Original Message ----- From: John F Hall
> To: Various people in computing, stats and quant method in sociology
> Sent: Saturday, August 07, 2010 9:33 AM
> Subject: SPSS, SAS and Stata
> 
> 
> I've read through the Mitchell article and it is more relevant to statististical aspects of surveys than to the sorts of things I cover in my tutorials.  It's very thorough, but there are no tables or figures showing direct comparisons of syntax or output from SPSS, SAS and Stata.  I'll add a link from my site, but I already have one to ATS at UCLA.
> 
> John Hall
> http://surveyresearch.weebly.com
> 
> 
> 
> ----- Original Message ----- From: John F Hall
> To: Reason Dave
> Cc: Various people in computing, stats and quant method in sociologySent: Friday, August 06, 2010 6:31 PM
> Subject: Sociologists and plumbers
> 
> 
> Dave
> 
> Just came across this opening paragraph in an article (link posted to SPSSX listserver by Dirk Enzmann)
> 
> There's huge debate going on about the relative merits of SPSS vs Stata and R, much of it an attack on the IBM/SPSS business model which is beginning to put SPSS out of the reach of many universities as well as students.  Many users are now switching to Stata, thus losing a whole generation of students and future users forever.  There are some heartbreaking, extensive, thoughtful and constructive contributions on the thread inexpensive 'home' version? started by Peter Neenan SPSSX-L@LISTSERV.UGA.EDU
> 
> Mitchell, M. N. (2005). Strategically using General Purpose Statistics Packages: A Look at Stata, SAS and
> SPSS (Technical Report Series, Report Number 1, Version Number 1). Statistical Consulting Group:
> UCLA Academic Technology Services. Available at http://www.ats.ucla.edu/stat/technicalreports/
> 
> This report describes my experiences using statistical packages over the last 20+ years, including my ex-
> periences as a statistical consultant at UCLA for more than 11 years. As a statistical consultant, I have
> worked with thousands of researchers and have worked with well over a dozen packages. In any given day,
> I bounce from helping people using Stata, then SAS, then SPSS, or Mplus, perhaps HLM, maybe LogXact,
> perhaps LatentGOLD, maybe MLwiN and so forth. I have seen how certain packages have certain strengths
> and others have certain weaknesses, and that these strengths and weaknesses fall along a large number of
> dimensions. I have come to believe that data analysts are like carpenters and that statistical software makes
> up the tools that we use. A carpenter would not buy a screwdriver and conclude that his or her toolkit
> is complete. Likewise, as data analysts, we may need to draw upon multiple tools (statistical packages) to
> form a complete toolkit based on the kind of work each of us performs.
> 
> The article was updated in 2007: he's still working at UCLA and is a regular contributor to Stata listerver.  I thought the carpenter analogy appropriate, given your remarks about plumbers and sociologists.  From the abstract and his acknowledgments, I detect a kindred spirit working in an environment not unlike SSRC and PNL.
> 
> John 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/