LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (September 2009)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Fri, 25 Sep 2009 23:05:10 +0200
Reply-To:     John F Hall <johnfhall@orange.fr>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         John F Hall <johnfhall@orange.fr>
Subject:      Re: Can we "freeze" at SPSS 15 for several years?
Comments: To: Mike Palij <mp26@nyu.edu>
Comments: cc: Westlake Andrew <ajw@sasc.co.uk>,
          Tony Coxon <coxont@Cardiff.ac.uk>, Jim Ring <ringj@btinternet.com>,
          Beverley Rowe <bev@bevrowe.info>, Banks Randy <randy@essex.ac.uk>
Content-Type: multipart/alternative;

Mike

Nice contribution: keeps the debate going. Had similar debates in UK in the 70s and 80s with statcomp nerds, but legitimate problems nevertheless. It's a bit late over here, so I'll check the citations tomorrow. Fortunately all my surveys were small enough to fit in SPSS. Did you ever see the manuals for BMDP and Osiris? Or Fakad? Yikes!! At PNL they used Minitab (double yikes!) but some people keep talking about SNAP. Must investigate.

John Hall ----- Original Message ----- From: Mike Palij To: John F Hall ; SPSSX-L@LISTSERV.UGA.EDU Cc: Mike Palij Sent: Friday, September 25, 2009 1:53 PM Subject: [Norton AntiSpam] Re: Can we "freeze" at SPSS 15 for several years?

If one wants "SPSS versus other software" one should look at the Tabachnick & Fidel "Using Multivariate Statistics" which is presently in the 5th edition. Available on Amazon and other booksellers, I suggest looking at the 4th edition which compares how SAS, SPSS, and SYSTAT (remember that?) compare in doing the same analyses: http://www.amazon.com/Using-Multivariate-Statistics-Barbara-Tabachnick/dp/0321056779/ref=sr_1_1?ie=UTF8&s=books&qid=1253876611&sr=8-1 or http://tinyurl.com/y8kxzxh

Back when I was in grad school (late 1970s, early 1980s) it was emphasized that doing any serious (high stakes) statistical analysis should be in at least two different statistical packages and to compare the output for both (I typically used BMDP and SPSS). This was a policy followed by some physicists who wanted to make sure that their home grown programs behaved appropriately even if then used different algorithms for calculations. Of more direct relevant to experimental psychologists like myself and social science researchers was the 1977 American Statistician article by Leland Wilkinson and Gerard Dallal which is available on www.jstor.org. Here is the citation info along with a stable Jstor URL:

a.. Accuracy of Sample Moments Calculations among Widely Used Statistical Programs b.. Leland Wilkinson and Gerard E. Dallal c.. The American Statistician, Vol. 31, No. 3 (Aug., 1977), pp. 128-131 (article consists of 4 pages) d.. Published by: American Statistical Association e.. Stable URL: http://www.jstor.org/stable/2682964 Wilkinson and Dallal showed how several different mainframe software programs (i.e., SPSS, BMDP, OSIRIS, and DATATEXT) could produce errorneous results because the programmers failed to take into account the possibility of underflow or overflow errors because they had not allocated sufficient space in memory to represent very small or very large numbers. Only BMDP, because it used a different algorithm was able to produce correct results. I wondered how many people who had used SPSS or the other programs for their dissertations or published researched actually went back to their data and re-analyzed it in order to find out whether they had made any errors. Subsequent versions corrected these problem but given the complexity of statistical programs, new problems will inevitably crop up.

Wilkinsion and Dallal would go on and develop various statistical testing batteries to see how accurate the calculations are for the newer statistical packages. B.D. McCullough has also done the same; see;

a.. Assessing the Reliability of Statistical Software: Part I a.. B. D. McCullough a.. The American Statistician, Vol. 52, No. 4 (Nov., 1998), pp. 358-366 a.. Published by: American Statistical Association a.. Stable URL: http://www.jstor.org/stable/2685442

a.. Assessing the Reliability of Statistical Software: Part II a.. B. D. McCullough a.. The American Statistician, Vol. 53, No. 2 (May, 1999), pp. 149-159

a.. Published by: American Statistical Association a.. Stable URL: http://www.jstor.org/stable/2685736

McCullough has also evaluated the statistical capabilities of off-the-shelf Excel and his recommendation is don't use Excel for any serious statistical analysis. You don't have to take his word

It should also be noted that Leland Wilkinson (his webpages is http://www.cs.uic.edu/~wilkinson/ ) developed the SYSTAT software package. It might be remembered that SPSS bought SPSS (along with a lot of other software packages) but then spun SYSTAT off (the story on this is provided by Wilkinson on his website; see: http://www.cs.uic.edu/~wilkinson/SYSTAT/systat.html )

So, SYSTAT is still available as a statistical package (I own a copy of version 11) and more information about it can be obtained from the SYSTAT website: www.systat.com I would suggest that people take it into consideration as a vaiable alternative to SPSS but such a decision would be dependent on how well SYSTAT meets you immediate and foreseeable needs.

-Mike Palij New York University mp26@nyu.edu

----- Original Message ----- From: John F Hall To: SPSSX-L@LISTSERV.UGA.EDU Sent: Friday, September 25, 2009 5:55 AM Subject: Re: Can we "freeze" at SPSS 15 for several years?

Looks fairly intuitive. Can it really do everything I want? If so, does this mean I have to ditch 38 years of SPSS?

My tutorials are built round the processing and analysis of actual data from real questionnaire surveys (no artificial data whatsoever) from initial checking of raw data to publishable tables and charts. I use SPSS because it was there when everyone needed it, was widely available and had a manual which made sense (even though I had to write tutorials to suit the research sequence rather than follow the alphetical sequence of commands).

I'm too busy doing my SPSS versions, but it would be interesting to see parallel STATA syntax alongside that for SPSS. More than 30 years ago, Steve Tagg and I thought of producing a Clods' Guide to Survey Analysis with SPSS, but we later also had the idea of a loose-leaf edition with side-by-side examples from other software. I don't know if anyone has actually produced such a manual, but if anyone wants to try it with any of my tutorials, feel free.

They're all listed on Survey Analysis Workshop (July 2009) with hyperlinks for free downloads. By the end of today there'll be even more uploads.

This thread is throwing up some really interesting comments, but I'm not sure the subject is catching everyone who could contribute. Perhaps something like, "SPSS versus other software"? ----- Original Message ----- From: Nils Braakmann To: John F Hall Cc: SPSSX-L@listserv.uga.edu Sent: Friday, September 25, 2009 11:00 AM Subject: Re: Can we "freeze" at SPSS 15 for several years?

I don't get the complete SPSS-synta, but I try my best. > recode V363 (1 3 5 7 =1)(2 4 6 8 =2) into ethnic. One possibility: recode V363 (1 3 5 7 =1)(2 4 6 8 =2) ren V363 ethnic

> rename variables (v348 = sex). ren V348 sex

> freq sex ethnic. tab sex ethnic (maybe with some suboptions, depending on what you want in each cell)

In fact, I don't think there's much of a difference between SPSS and Stata for these simple commands. The main advantage of the Stata-syntax is that it always follows the structure "command variables, options" which makes it pretty intuitive. For instance,

Linear regression reg y x

with robust standard errors reg y x, robust

with bootstrapped SEs (500 reps) reg y x, vce(bootstrap, reps(500))

Logit logit y x

with robust standard errors logit y x, robust

And so on.

Best,

Nils On Thu, Sep 24, 2009 at 10:58 PM, John F Hall <johnfhall@orange.fr> wrote: > If SPSS syntax is: > > recode V363 (1 3 5 7 =1)(2 4 6 8 =2) into ethnic. > rename variables (v348 = sex). > count sexism = V248 V252 V253 V256 V261 (3,4) > > V251 V255 V259 V260 (1,2). > > freq sex ethnic. > freq sexism /his nor. > cros sexism by sex ethnic > /sexism by sex by ethnic /cel per. > means sexism by sex ethnic > /sexism by sex by ethnic. > > ...what would the syntax be in STATA and would it be as easy to understand? > >


[text/html]


Back to: Top of message | Previous page | Main SPSSX-L page