Discussion: View Thread

Is SAS on the decline?

  • 1.  Is SAS on the decline?

    Posted 01-29-2014 10:29
    One of my students sent me a blog entry that claims SAS and SPSS are being used much less often than before and will eventually give way to R. I would like to know what people on this list think of this. It seems timely in light of our recent discussion of lower-cost alternatives to SAS.

    As a program evaluator and professor of psychology, I use SPSS because it's still standard in our field, required by my department for our courses, and we don't need to do the kinds of programming that people in this group do. So I defer to the expertise of those in this group.

    Here's the article; it's not very long, but does include some graphs to support its argument:
    http://r4stats.com/2012/05/09/beginning-of-the-end/

    Also, I would appreciate if anyone can recommend a book or resource for someone learning R for the first time.

    Thanks!

    -------------------------------------------
    Annette Gourgey, CUNY
    -------------------------------------------


  • 2.  RE:Is SAS on the decline?

    Posted 01-29-2014 10:43
    If you are learning R and know SPSS (or SAS) I recommend R for SAS and SPSS users by Robert Muenchen

    As for "declnes". Well, we will see.

    -------------------------------------------
    Peter Flom
    -------------------------------------------








  • 3.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:00
    Thanks for the interesting article Annette.  I personally prefer R but I am currently working on clinical trials and I think that as long as the FDA requires SAS for clinical trials, it's not going away anytime soon. 

    The price of R can't be beat and on a regular basis I get questions for colleagues on how to learn R because many institutions don't want pay for SAS and SPSS.

    In addition to the book that Peter mentioned below, one of my favorite go-to books for teaching R and my own reference is 'R in Action' by Robert Kabacoff.  Kabacoff is currently working on a second edition, but the first is very good.  He also maintains an R help website, Quick-R.

    -Penelope 

    -------------------------------------------
    Penelope Pooler
    Statistician
    SUNY Upstate Medical University
    Dept. of Public Health and Preventive Medicine
    -------------------------------------------








  • 4.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:08
    I don't do clinical trials or work with the FDA but I found this which says the FDA does allow people to use R.

    I can't say anything about that from experience, just posting it for information

    -------------------------------------------
    Peter Flom
    -------------------------------------------








  • 5.  RE:Is SAS on the decline?

    Posted 01-29-2014 18:27
    The purchase price for R is wonderful.

    However, the total cost of analysis for R  is much higher than for SPSS hence also for the more expensive SAS.

    <tongue in cheek>
    Of course the value of student time is zero>
    <\tongue in cheek>

    YMMV but a subjective guess from my experience is that is the data prep and cleaning can easitly exceed 90% of the staff time on a project.
    SPSS has a great lead in this part of an analysis.  SAS is not quite as oriented to human factors as SPSS but is much better than R for doing this part.  The messier the data the greater the advantage of SPSS.

    R is very ad hoc for different procedures.  If one is going to be doing pretty much the same thing in one' research that does not make much difference. If one is going to be doing a wide variety of things then SPSS and SAS are larger tool boxes.

    There is an urban legend that FDA routinely violates regulations by insisting on SAS but that urban legend is widely held and it can be easier to go along to get along.  Much consulting work is outside the pharmaceutical area so what some believe about FDA is irrelvant in those areas

    Quality assurance in an analysis is much more efficient the more readable the syntax used. Ditto maintenance of procedures that will be run routinely.

    For beginners readability of syntax and of output greatly facilitates understanding of what is done in an analysis.


    Businesses are increasingly using the kinds of stat from the social sciences. Much of business culture has a total cost vs purchase cost outlook, so I think integrated packages will gow in use in 'business intelligence".

    Some of the output from R has wording in it that gives the impression that the procedures are still under development.  Attorneys for the side that did not do he analysis will make a big point of that for the trier of fact whether the trier of fact is a judge or a jury. 

    SPSS users can use SPSS for all of the meta data and as long as the dataset is small enough to fit in RAM R procedures can be called without dealing with the R  data handling.   I heard someone say that SAS was considering doing the same thing.

    Some people who teach stat these days (I have retired) and who have been forced to change from SPSS or SAS to using R have told me that R makes it harder to students to relate what they are doing to the substance of the analysis. I.



    -------------------------------------------
    Arthur Kendall
    Social Research Consultants
    -------------------------------------------








  • 6.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:55

    Another excellent textbook that I would recommend for R-SAS users is "SAS and R" by Kleinman and Horton.


    -------------------------------------------
    Elena Randou (Rantou)
    FDA
    -------------------------------------------








  • 7.  RE:Is SAS on the decline?

    Posted 01-29-2014 10:46
    Annette, you may want to check out Stanford's free online course "Intro to Stat Learning" with its free PDF book by the instructors:

    http://online.stanford.edu/course/statistical-learning-winter-2014

    -------------------------------------------
    Wayne Fischer
    Statistician
    University of Texas Medical Branch
    -------------------------------------------








  • 8.  RE:Is SAS on the decline?

    Posted 01-29-2014 10:46
    That blog article looked specifically at trends in scholarly applications.  True, R is likely to become more prevalent is academia and small businesses (where "free" is highly attractive).  However, SAS will not go away in business and industry, especially in healthcare fields.  The high price tag for SAS does provide a big return in productivity.

    -------------------------------------------
    Mark Martin
    Siemens Healthcare Diagnostics
    -------------------------------------------





  • 9.  RE:Is SAS on the decline?

    Posted 01-29-2014 10:52
    That's all very interesting, and I think that SAS is going to be a little diminished. But the notion that 2015 will be the "end of SPSS and SAS"? Doubtful, to say the least.

    1) SAS and SPSS have corporate resources which ensure that the products are back-ward compatible. R does not. Some R developers have stated that they do not spend their time ensuring that they get the same results as SAS does. OK, good for them. I'm not comfortable using a product which gets different results and is unwilling to address reasons. SAS has staff which ensure that the product does what it is supposed to do, and is backward compatible.

    2) The FDA is the firewall for SAS. When the FDA accepts a submission with an R analysis, that will be a signal moment. To the best of my understanding, analyses submitted to the FDA require specific products.

    3) Many R products will not produce p values. Fine by me, but I am interested in p values, and don't want to spend endless hours in annoying discussion of why my interest in p values makes me a less competent statistician. When they produce p values and let the user decide which parts of the output to report, that would be also a good thing.

    4) SAS is too expensive for a lot of people. It used to be that SAS was cheap/free for students. Now R is free, and SAS has a cost for students (that may not be right, but I'm not a student or a teacher at this time). When I look for MA staff, I look for those with SAS and R.  That's what I see happening for the next 5-10 years. By 2020, dunno - things may be very different.

    5) I'd be interested in what is going on in Europe, especially the smaller less wealthy parts, and in the non-west - South America, India, Africa. Many of these countries have less wealthy users, and may not be willing to learn SAS and pay the fees.



    -------------------------------------------
    Paul Thompson
    Director, Methodology and Data Analysis Center
    Sanford Research/USD
    -------------------------------------------








  • 10.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:06
    To add to the fun, we shouldn't forget the enterprise modifications of R done by Revolution Analytics, which has the potential to be an alternative to SAS's enterprise.

    The bottom line for many customers is cost, and R is very affordable.

    As for the whole SAS and the FDA stuff, I'll just leave this link here...
    http://blog.revolutionanalytics.com/2012/06/fda-r-ok.html

    -------------------------------------------
    Mark Lancaster
    George Mason University
    -------------------------------------------








  • 11.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:15
    As someone currently teaching in Latin America, I would say for us, the cost of software is an issue. I teaching the majority of our statistics-related courses in the School of Agronomy, and for many they have become accustomed to software from Argentina that is quite good especially when considering the language, as well as software that is connected to R for analyses like mixed models. Nonetheless even this software comes at a cost, although compared to SAS or SPSS, for example, at a much reduced price. Given all of that then, for the graduate level courses, I teach primarily in R and RStudio. Working with colleagues, many of whom studied abroad, SAS is a program still of much interest based on their experiences and I continue to use this for specific analyses where I am maybe more familiar with the programming and interpretation. As for the future, I think realistically SAS may not be available in a few years for us, somewhat based on the University-level desire to move as much as possible to freeware. 

    -------------------------------------------
    Paul Esker
    Professor
    Universidad De Costa Rica
    -------------------------------------------








  • 12.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:34
    Very interesting to say the least.

    I'm a long-term SAS user, and don't use R much, if at all. People like me are the reason SAS is still around. When people who have trained with R primarily and SAS as a secondary tool get more into management and the responsibility to select tools, you can expect changes. Many persons who are at my level have already switched to R, and either don't use SAS or actively make snide comments about it. So, by 2025, in 10-11 years, I dunno what things will look like but I do expect that SAS will be a smaller part of the pie.

    -------------------------------------------
    Paul Thompson
    Director, Methodology and Data Analysis Center
    Sanford Research/USD
    -------------------------------------------








  • 13.  RE:Is SAS on the decline?

    Posted 01-29-2014 13:00
    Hi all,

    Very interesting thread, I have some friends who work at SAS who will be interested (and amused) to read it.

    I think we have to separate our discussions about SAS the software and SAS the company.  SAS the company has started providing a wide range of software in addition to the original code based SAS.  They consult with companies for analytic, data management, business intelligence, and high speed computing solutions that are individually client based.  The company is not in decline and the only reason why you don't hear about them all the time on MSNBC is that they don't have stock so they aren't interesting to market analysts.

    Now as for SAS the software, I wouldn't say it is in decline either.  I agree with the earlier comments on SAS code with FDA submissions as well as software costs and R as an attractive alternative.  But keep in mind that SAS is also a very powerful Data Management software.  We statisticians tend to think of ourselves as THE source for all things data but many tend to work only in analysis and not in data management, manipulation, or storage and processing.  I recently met a group of data technicians that work for a state-level insurance agency that spend all day programming in just Base SAS.  Imagine using SAS all day but with no stat packages.  If you attend SAS user group meetings you will find as many talks (maybe more) about processing and managing data as you will find tips and tricks for analysis. I would say that the existence of R has been a boon to SAS because R is certainly a good playground for academics and others who want to push the envelope on statistical analysis in an open format.  SAS has more time and options for evaluating new methods to implement in updated versions of SAS without people always bugging them about adding the next big thing.  SAS has developed good R integration (just like it did with SUDAAN and is now doing with HADOOP); I think the misconception is that it has to be one code based software or another.  

    There are other good stats software on the market, STATA has some ridiculously cool stuff.  JMP is one of my favorites because of it's ability to interact with data, I use it probably 3 days a work week.  NCSS, Minitab, and SPSS are all ok but tend to be limited in some of their options or processing capability. But if you are going to compare anything to SAS then make sure you ask about how well it runs on multiple processing cores, or how it handles data management, as well as versatile it is for statistics and programming.

    Many R users I know are very anti-SAS but many SAS users I know are not anti-R.  If you want a good resource on both check out the SAS and R blog (http://sas-and-r.blogspot.com/) which is one of my favorites.  I am a regular SAS user and often get frustrated in R because I have so much more experience with SAS, this blog is good because it often times  shows you how to do the same things in both software, helping to build a bridge.

    -------------------------------------------
    Jason Brinkley
    East Carolina University
    -------------------------------------------








  • 14.  RE:Is SAS on the decline?

    Posted 01-29-2014 13:35
    I run a very small shop, and while I cringe every year when the invoice from the SAS Institute arrives, I invariably pay it.  Why?  Because of the productivity gained and the resultant increase in billable hours.  I am learning R, and in time perhaps I will use it more.  

    I also have some fond customer service memories dating back many years that gives the SAS Institute some credibility as a vendor that the diffuse structure and development associated with the R venture is not, at least at present, able to replicate.

    It is fascinating, however, to see the comments associated with this evaluation of tools.


    -------------------------------------------
    David Mangen
    -------------------------------------------








  • 15.  RE:Is SAS on the decline?

    Posted 01-29-2014 13:37
    This is an interesting discussion for me.  I'm also a long time SAS user and would prefer to keep using SAS but the cost is becoming very expensive.  I have been considering trying R.  Thanks for the references.

    -------------------------------------------
    Rocco Brunelle
    Senior Statistician
    Bowsher Brunelle Smith LLC
    -------------------------------------------








  • 16.  RE:Is SAS on the decline?

    Posted 01-29-2014 14:25

    I understand that the main issue between commercial vs. 'opensource', validation software is 'validation'.

    In this context, 'validation' means acceptability of results from the point of view of reproduceability.

    Commercial software requires months/years of testing to 'assure' users that at all times the same code/syntax will produce the same results from the same dataset.

    'Opensource' should havetgran the same capability, but is not necessarily the case.

    Thus, it seems reviewers of grants and or of publications reject proposal and/or results generated from other than commercial software.

    Please, comment!!

    -------------------------------------------
    Anamaria Kazanis]Designation]
    Senior Statisitican
    Michigan State University
    -------------------------------------------








  • 17.  RE:Is SAS on the decline?

    Posted 01-29-2014 14:45
    I see a few issues:

    1) Certainly journals accept articles written with analysis done in R.  I can't say there are NO journals that do NOT, but there have been many articles published based on R analyses. The same is true for grants.

    2) The core packages of R are, by now, very well tested indeed.

    3) As someone else pointed out, the code in R is visible. So, experts can look at it and suggest corrections. This, of course, does not assure that there are no bugs. But no software is completely bug free. Commercial stats packages are, of course, tested. In the case of SAS (the only one I am even vaguely familiar with) I am pretty sure they are very well tested. But does that mean they have no bugs? Well.... How can one ever be absolutely sure?

    4) Given how many errors people make with the code, the data and the choice of analysis - well, I am guessing that software bugs are probably far down the list of things that cause errors in articles.

    5) Where I see the big advantage of SAS, within the statistics realm, is in the excellent documentation and technical support.  In my experience, package developers in R are usually also helpful, but not to the same extent; nor would I expect them to be - SAS pays people to help users. R doesn't. There are many good books on SAS and on R; but I find the built in help in SAS excellent. R prides itself on a terse style of help that I find less helpful. Others may differ. Try typing ?median in R.

    Peter







  • 18.  RE:Is SAS on the decline?

    Posted 01-29-2014 15:26
    At the risk of sounding like a shill, one of the things that does impress me about SAS is their transparency regarding bugs or differences of opinion regarding how some matters should be computed.  I receive a daily email that documents these bug reports, as well as notifications regarding how to correct/apply updates to solve the problem(s).


    -------------------------------------------
    David Mangen
    -------------------------------------------








  • 19.  RE:Is SAS on the decline?

    Posted 01-29-2014 18:20
    Hi,

    I am a multi-lingual statistical software user, having started with SAS in 1986 and then learning SPSS, LISREL, AMOS, Mplus, and Stata, depending on the academic environment in which I've been working and the requirements of the various projects I've been involved with. I've appreciated seeing the references folks have posted for R for users of other programs because I have not found R to be intuitive the few times I've tried to pick it up (which I don't blame on R, just my own thick-headedness in trying to learn object-oriented programming).

    Following up on Peter's 5th point below, I, too, have found SAS tech support be excellent over the years I've used SAS. To that I would add how incredibly beneficial I've found the in-person and online courses I've taken from SAS Education. I've learned so much about applied statistical data analysis from those courses and highly recommend them for SAS users who are interested in the topics covered by the classes. The availability of a wide variety of SAS Education courses and their excellent instructors have helped me stay up-to-date with the various analytic methods covered in the courses and these courses' availability is one of the things that keep me returning to SAS for many of my statistical computing needs.

    With best wishes,

    Tor Neilands

    -------------------------------------------
    Torsten Neilands
    Professor
    UCSF Center for AIDS Prevention Studies
    -------------------------------------------








  • 20.  RE:Is SAS on the decline?

    Posted 01-29-2014 14:38
    Someone mentioned the other day a generic version of SAS (i.e., reverse engineered to use the same syntax) called The World Programming system.
    http://www.teamwpc.co.uk/products/wps

    They quoted me an annual license fee of about $1,400, however their stat package is a basic subset of SAS stats.  I don't need SAS at home, so I haven't pursued it any further, but someone else might be interested

    -------------------------------------------
    Dennis Sweitzer
    Principal Biostatistician
    Medidata Solutions
    -------------------------------------------




  • 21.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:22
    Backward compatibility (especially for reproducibility of the results) is not a trivial matter from the regulatory perspective. Applies to parts of financial services as well.

     - Michiko

    -------------------------------------------
    Michiko Wolcott
    Lead Data Scientist
    Enfathom, a Division of North Highland
    -------------------------------------------








  • 22.  RE:Is SAS on the decline?

    Posted 01-29-2014 15:37
    There are a few issues that need correcting here.

    1. SAS is not completely backwards-compatible. One of my colleagues has an example of this. The same code run on different versions of SAS produces different results. It may be mostly backward-compatible, of course, and it may be MORE backwards-compatible than R, though I'm not sure how you would test this. Also, making sure that R produces the same results as SAS is not the same thing as making sure that R produces the correct results. I know that the R core development team spends a lot of time on the latter.

    The point that Dr. Thompson is making here is that SAS has a lot more money coming in than R and therefore it spends a lot more money on quality control than R does. This is an attractive argument, for SAS, though there are some counter-arguments. Open source programs like R are more transparent and thus more amenable to quality control efforts by outsiders than closed systems like SAS. Does anyone here know how SAS computes Type IV sums of squares? Does anyone here know if SAS computes them correctly?

    I suspect that the quality of the core R functions is comparable to the quality of the comparable SAS functions, but that is a conjecture. Certainly one reason to pay more for SAS is to pay for the extra quality control that SAS gives you. SAS is not perfect, but it is probably pretty close, given the extensive resources that they spend on their product. It is one of the strongest arguments in favor of SAS, in my humble opinion.

    2. I don't know the first time that FDA accepted a submission with an R analysis, but it has indeed happened. In fact, R is used a lot by FDA statisticians themselves. An article published seven years ago made this point very well.
    --> http://user2007.org/program/presentations/soukup.pdf

    3. I'm a bit mystified by this comment. Certainly t.test produces a p-value. lm produces a p-value, though you have to use summary to get it. I am having a hard time thinking of a statistical procedure that uses p-values and which you can't get the p-values out in R. Can you give an example? Furthermore, you can store the output of any R function and extract what you want. In fact it is far easier to do this than it is in other packages. Just as one example, how easy is it to round the p-values produced by SAS in glm or reg to a different number of digits? I know it can be done, but it is not trivial. In R, you run summary and store it and pick out the p-value (I think it is the 2nd row, 4th column of the coefficients table). I do that all the time. You can even poke the rounded p-value back into the original object (a frightening thought, actually, is that R would allow you to poke a DIFFERENT p-value back in). Anyway, customizing the output in R is one thing that R does BETTER than most other statistical packages.

    It's easy to come across as antagonistic in these emails. Although I do want to "set the record straight" that does not mean that I disagree with the overall thesis of Dr. Thompson's post. SAS has a lot going for it and I would never criticize someone for preferring SAS. You do pay a lot for SAS, but you get a lot in return. Also, the cost of R may be less if you look just at the initial purchase price, but when you factor in other costs (such as support), it may be more, especially in a big organization. There's no easy answer to the question of what statistical software is the best.

    -------------------------------------------
    Stephen Simon
    Independent Statistical Consultant
    P. Mean Consulting
    -------------------------------------------








  • 23.  RE:Is SAS on the decline?

    Posted 01-29-2014 16:15
    I hope that no one is getting confused by these comments. I am sure that R does a lot of things correctly. Many seem to get a lot of benefit from it.

    1) Some say that the ability to read the actual code in R is an advantage. OK, let's pretend that it is. Who here is going to open up lmer or some other COMPLEX R function and tell us all that you will be able to detect if the code is correct or not by looking at it? Having spent more than my fair share of time on "forensic code degobblification", I find such notions implausible, to say the least. Just because you CAN look at the code does not mean that you think, "Gee, I need to run a mixed model. I'll begin by reviewing the code." So, VIEWABLE code is not the same as TRANSPARENTLY UNDERSTANDABLE code. In many cases, code is hard to understand, even for the author (after a period of time).

    2) On the p-value thing - I was speaking of lmer/lme4, specifically, where the author has stated that he won't provide p values.

    Users are often surprised and alarmed that the summary of a linear mixed model fit by lmer provides estimates of the fixed-effects parameters, standard errors for these parameters and a t-ratio but no p-values.  Similarly the output from anova applied to a single lmer model provides the sequential sums of squares for the terms in the fixed-effects specification and the corresponding numerator degrees of freedom but no denominator degrees of freedom and, again, no p-values.

    https://stat.ethz.ch/pipermail/r-help/2006-May/094765.html

    and there are many other discussions with this point, and others like it.

    Fine by me. Works for many. I'd prefer a more flexible approach to df-denom. In SAS, you select your approach, and it computes a p value. In R, you apparently have to use a more complicated approach. I'm sitting here thinking of how I would examine the tests associated with a relatively small trial. I'm unsure as to why you can't simply define the df approach and have lmer or lme4 or whatever do the calculations. 

    -------------------------------------------
    Paul Thompson
    Director, Methodology and Data Analysis Center
    Sanford Research/USD
    -------------------------------------------








  • 24.  RE:Is SAS on the decline?

    Posted 01-29-2014 16:43
    One other point:


    Steve Simon commented:
    QUOTE

    In fact it is far easier to do this than it is in other packages. Just as one example, how easy is it to round the p-values produced by SAS in glm or reg to a different number of digits? I know it can be done, but it is not trivial. In R, you run summary and store it and pick out the p-value (I think it is the 2nd row, 4th column of the coefficients table). I do that all the time. You can even poke the rounded p-value back into the original object (a frightening thought, actually, is that R would allow you to poke a DIFFERENT p-value back in). Anyway, customizing the output in R is one thing that R does BETTER than most other statistical packages.

    ENDQUOTE

    The issue of formatting of output is one thing that SAS has really improved in the last couple years. Producing p values to different numbers of digits is easy -

    4 digits: format pvaluevar pvalue6.4;
    2 digits: format pvaluevar pvalue6.2;

    so that's pretty straightforward.

    As to other structures and printing of output, I look at what SAS produces directly for myself, but I never show that to end users.  Every output table that SAS produces can be placed into a dataset, and then that dataset can be printed in a better, more readable, and more easily understood manner. SAS has greatly improved this. I'm sure R does this well too.



    -------------------------------------------
    Paul Thompson
    Director, Methodology and Data Analysis Center
    Sanford Research/USD
    -------------------------------------------








  • 25.  RE:Is SAS on the decline?

    Posted 01-29-2014 17:15
    A great point!
    Probably for any SAS procedure you can output just about any element of results using ODS. For example, the list of ODS table names for PROC REG:
    http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_reg_sect049.htm
    These tables are simply datasets that could be combined in a data step, output as excel file, etc. This is especially handy when you want to compare results from different models.

    With respect to graphics (replying to Susan's post here), SAS has also improved greatly. I agree SAS Graph is cumbersome and not very intuitive. Currently, I'm fond of ODS graphics. This book by Sanjay Matange is a good reference "Statistical Graphics Procedures by Example".


    -------------------------------------------
    Yelena Perkhounkova
    Statistician
    University of Iowa
    -------------------------------------------








  • 26.  RE:Is SAS on the decline?

    Posted 01-29-2014 17:49
    I totally agree with the recommendation of Sanjay Matange's book "Statistical Graphics Procedures by Example".

    I just reviewed this book for The American Statistician (*insert shameless self plug*) for the May 2014 issue.  It really is a good book with a ton of ready to use code.  I especially liked the presented code for Meta Analysis Forest Plots and for Adverse Event Timelines in drug development.  Also you can get it in paperback for about 50 bucks so if you have a small book or supplies budget then you can easily work it in.

    My only complaint is that they printed it in greyscale (to keep costs low?).

    -------------------------------------------
    Jason Brinkley
    East Carolina University
    -------------------------------------------






  • 27.  RE:Is SAS on the decline?

    Posted 01-29-2014 18:20
    This thread and the earlier one on substitutes for SAS has been most interesting.

    I'm old enough to have first used SAS on punchcards.  More than 20 years ago I first started using JMP, I was attracted to the ease in making and modifying graphs.  It was only available on a Mac at that time but that was a platform available to me.  So I used SAS for maybe 90% of my work and JMP for 10%.  And since JMP is made by SAS it was easy enough to move data back and forth. 

    Over time JMP has continued to add functionality and expand its capabilities such as the ability to handle larger and larger data files.  And it became available on the PC in the mid 90s.  The 90/10 ratio of SAS to JMP eventually flipped to 10/90 and then the 10 contuinued to get smaller and smaller (I'll avoid my itch to make a 1% joke here!).  As our group added new members many had experience with (and sometimes a preference for) R, so we made it available. 

    As the need for SAS ultimately became very small we discontinued the licenses, and for the very few things that we cannot do in JMP we now use R.  Interesting that one product's expansion made us abandon the other as they both come from the same company.

    I should also mention that Minitab is also used extensively in our organization though I find it has fewer capabilities than JMP.


    -------------------------------------------
    Francis Rossi
    Associate Director, Statistics
    Kraft Foods Group
    -------------------------------------------








  • 28.  RE:Is SAS on the decline?

    Posted 01-29-2014 17:00
    I won't comment on the subject line specifically, as I know SAS has many areas of business beyond the core products of Base SAS, SAS Statistics, SAS Graph, etc.  

    I am currently working in the area of benefit-risk, however in the past I've had interest/involvement in Statistical Computing Environments and statistical graphics (including a business lead role in defining requirements, identifying software, and deploying a statistical graphics solution at my company for statisticians' use).  

    Stephen Simon mentions an FDA reviewers presentation on this topic.  I had not seen this presentation until now, but have discussed with that reviewer and a few others at FDA.  His presentation brings up two thoughts:

    1. From a regulatory standpoint, there are both regulatory reviewers and compliance auditors (for 21CFR Part11 compliance).  Mat addressed both in his presentation.  However it is one thing to convince the FDA stat reviewer/FDA auditor that use of R is acceptable.  It's quite another to convince IT and compliance people at one's own organization that use of an open source software is acceptable.

    Whilst I've found an openness to the possibility with IT and compliance colleagues, I can understand their viewpoint too.  SAS licenses may cost more, but probably most organizations would either use only SAS or both SAS and R (ie, not R alone).  IT and compliance people see these tools as just two more to manage.  We've had to explain why have two software packages when one will do?  

    As a revered (and now retired) senior stat manager once told me, our role as statisticians in R&D is to push the envelope - to find new and better ways to do our work.  For IT and Compliance, their role is to manage risk (and for IT, cost as well, which includes internal costs of managing/updating software).  It's my belief that these differing roles/responsibilities within our own organizations is the larger barrier to uptake of R in FDA submissions.  

    2. The other thought Mat's presentation raises is statistical graphics.  The ease with which one can explore graphics in SAS is, well, suboptimal, although they continue to improve the product.  JMP is better (but again, that's another package).  

    An example: there are a multitude of drivers for SAS Graph, and only some of them work well for graphics.  Within my company we sorted out which ones worked with hardware fonts (our regulatory group was complaining the graphs with soft fonts made the files very large).  In S-Plus, they simply use a good driver for all the graphics - it's not even a user option to deal with.

    Back to the presentation, it points out that graphics is quite important in safety.  Software that lowers the barrier to creation of well-designed (safety) graphs is a plus.  (and here I'll shamelessly plug the FDA/Industry/Academia Safety Graphics Working Group's wiki: https://www.ctspedia.org/CTSpedia/StatGraphHome - it has code in SAS *and* R for well-designed graphs that answer commonly asked safety questions.)

    ---
    I agree with a previous comment that SAS is great tool for reshaping the data.

    -------------------------------------------
    Susan Duke
    GlaxoSmithKline
    -------------------------------------------








  • 29.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:16

    Annette,
    Thanks for sending the superb article.  I often recommend R, but I also often continue to use SPSS because I am familiar with it and its GUI and I'm not much of a programmer.

    There are many books on statistics with R, but most of them are devoted to elementary statistics (which readers of this list don't need). Such books tend to be "about R" only in the sense that the examples are worked out in R. 

    Likely to be much better for readers of this list is the book by Zuur, Ieno, & Meesters, A Beginner's Guide to R (Springer, 2009).  It teaches you the programming you need to use R, not how to do t-tests, ANOVA, correlations, etc., and the programming instruction is very elementary.
    Sincerely,
    Paul Vogt
    -------------------------------------------
    W. Vogt
    Professor
    -------------------------------------------








  • 30.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:19
    For good or ill, SAS is ubiquitous in government.  There is far too much legacy code in state and federal agencies to be going away anytime soon, not to mention heavy investments in SAS training.  Unless there is an R module that can execute complex SAS code without modification (please email me about that!!), nothing less would be remotely cost-effective.

    Many epidemiologists prefer STATA or SPSS for their own projects, not least because they learned them in grad. school.  R is making some inroads among individuals with the determination to learn, mostly on their own--I think training is encouraged but not often funded.  I have been working on it, but it's hard to find the time between projects.

    I have found R for SAS and SPSS Users and Data Manipulation with R to be helpful.  I have borrowed both from libraries, but I think there is a free PDF version of Data Manipulation with R, based on an early edition.

    Dave
    -------------------------------------------
    David G. McBride
    Senior Epidemiology Specialist, Office of Epidemiology
    Missouri Department of Health and Senior Services
    -------------------------------------------








  • 31.  RE:Is SAS on the decline?

    Posted 01-29-2014 11:53
    There are no specific requirements by FDA or any regulatory agency that require the use of SAS.

    I use both R and SAS. Anything I send to a regulatory agency is in SAS. exactly one time, I sent some analyses to FDA in SPLUS. 
    Some of the reasons for using SAS are, there are many expert SAS programmers, and many Pharma companies have developed libraries of SAS programs/macros for generating analyses. R programmers tend to be difficult to find.
    SAS is also "21 CFR part II" compliant, a regulatory requirement, and R does not yet have that status.

    My narrow experience outside the U.S., when I gave some lectures at the medical school in Fiji, they had SPSS. In Cuba, the statisticians used R.



    -------------------------------------------
    Chris Barker, Ph.D.
    Consultant and
    Adjunct Associate Professor of Biostatistics
    www,barkerstats.com

    ---
    "In composition you have all the time you want to decide what to say in 15 seconds, in improvisation you have 15 seconds."
    -Steve Lacy
    -------------------------------------------








  • 32.  RE:Is SAS on the decline?

    Posted 01-29-2014 12:35
    I serve clients in the financial sector. They are happy to pay me to replace SAS code with R code. It's a huge win for them. Within a few months of canceling their SAS license, they have recouped the cost of conversion... despite my outrageous fees.

    As an extra added bonus, they can immediately begin running their large applications in parallel across multiple machines, slashing their processing times without paying additional licencing fees.

    I have yet to encounter any SAS processing that could not be replaced with R. Learning R is not a hurdle. And the banking regulators are OK with the conversion.

    Will R replace SAS completely? Of course not. But my customers, at least, have no reason for keeping SAS.

    -------------------------------------------
    Paul Teetor
    Quantitative Developer
    Elgin, IL United States
    -------------------------------------------








  • 33.  RE:Is SAS on the decline?

    Posted 01-29-2014 16:38
    This is the first time I write to this group, so I hope I do it correctly.
    It's a very interesting discussion. I admire people who use R and thought of using it myself. I even took a couple workshops to make it easier to start. However, in the end my conclusion was that SAS can do everything I need and the time is limited to invest into another statistical package. I've used SAS for many years, but began to really appreciate it with SAS Enterprise Guide that made things much easier for me. The point-and-click interface made a huge difference for my new data manager and RAs who had no programming experience at all. They could be productive almost immediately when they started, and were able to learn how to write code gradually as data management needs dictated. SAS EG has some really cool features. You can still write code and use any procedures, but basic data exploration and simple stats are much easier with ready tasks. What I like especially is the ability to work in a project environment that has everything (data, programs, outputs, logs) in one place, showing connections among those elements. It's easy to locate things, and easy to work as a team when several people can access the project. That's why I'll be using SAS as long as I don't have to pay for it!  


    -------------------------------------------
    Yelena Perkhounkova
    Statistician
    University of Iowa
    -------------------------------------------








  • 34.  RE:Is SAS on the decline?

    Posted 01-29-2014 21:51

    What a provocative subject Is SAS on the decline?  What is the evidence presented by Annettee's students?  It's a webpage at r4stats.com.  Their mission is to analyze the world of analytics, and to help people learn to use R.  In other words, the author is not unbiased in this question.

    The basis for the claim is the number of hits reported by Google Scholar for each of the major software packages.  Their data is also available for download, but what is this data?  It's not the number of hits, in the sense used by Google Trends, it's the number of scholarly articles that reference the software.  As stated at scholar.google.com, Google Scholar provides a simple way to broadly search for scholarly literature.

    The basis for a predicted decline in SAS and rise in R is based upon scholarly articles that reference the software.  It's not based on the number of users and certainly not on software sales.  I believe Darrell Huff talks about this in Chapter 7  - The Semi-attached Figure.   Huff advises If you can't prove what you want to prove, demonstrate something else and pretend that they are the same thing. (pg 72 of How to lie with statistics)

    What does matter?  Well last year, SAS revenue topped $3B for the first time in their history.  This was their 38th consecutive year of growth!  I don't know of another company that can make that claim.

    What about R?  We might compare SAS against the growth in R users.  I looked at the growth in another remarkable company - Revolution R.   Formed in 2007, Revolution R has been able to raise over $30M.  In 2012 their revenues increased to $13M of which 44% was from services.  They too are growing.

    My point is that the statistical argument for a decline in SAS relative to R is unsupported.  In fact 38 years of positive growth indicates that SAS is not declining; to the contrary it's growing rapidly keeping pace with the increased use of Analytics in research and industry.  Revolution R is also growing but not at the expense of SAS.  They are both growing.

    Now if we really wanted to "statisticulate" about R vs. SAS, look at the following Google Trends graph.  It compares the number of google searches in the U.S. for "R software" vs "SAS software" from 2004 to the present.

    With this maybe we can declare R won the battle in 2007; the same year Revolution R was formed.  Yes, it's another semi-attached figure ;-).

       
    -------------------------------------------
    Edward Jones
    Executive Professor
    Texas A&M Univ.
    -------------------------------------------








  • 35.  RE:Is SAS on the decline?

    Posted 01-30-2014 06:45
    Anette,
    as a psychology professor you might know the SPSS-using textbook of Andy Fields. He has married an R-user and their child is ..
    http://books.google.de/books/about/Discovering_Statistics_Using_R.html?id=wd2K2zC3swIC&redir_esc=y

    -------------------------------------------
    Reinhard Vonthein
    Universitaet Zu Luebeck
    -------------------------------------------








  • 36.  RE:Is SAS on the decline?

    Posted 01-30-2014 10:43
    What a great discussion thread!

    I am a statistical consultant and my client are mostly pharma.  I used SAS 99.9% of the time, but I have also used other tools, such as R.  To me the question is more about understanding the HOW  vs the WHAT.  By HOW I mean, the mechanics (mathematics) behind the analyses as opposed to the WHAT, the interpretation of the statistical outcome.  When I use R, I feel like I am writing very basic code to do things that are already modulated in SAS.  It is like running a regression using matrix algebra.  I know that is what is going on in the background, but I want a tool that allows me to do it without all the coding.  Let someone else write the background code and validate it so when I call that module, I know I will get consistent and reliable answers.  R is moving in that direction and I can see it as a great teaching tool to show students the mechanics/mathematics behind the statistical approaches.  But as a consultant, I am willing to pay for the luxury of having someone else to do the background work so I can utilize the tools to provide what my clients want...the answers.

    -------------------------------------------
    Susan Spruill
    Statistical Consultant
    -------------------------------------------








  • 37.  RE:Is SAS on the decline?

    Posted 01-30-2014 11:10
    I've been reading this series of messages, and to myself (a SPSS programmer who joined a SAS government agency then a SPSS agency, then retired to tutor in stat and SPSS), Susan's message below hits the nail on the head. She and several others have stressed that reliability, not just ease of programming, matters a great deal. What language/series of commands/macros you can use to express the underlying logic and thus create an accurate statistical result is fine.  If I can clearly think through the underlying logic of what I want to achieve, then I can switch to the language I need to create that product.

    (By the way, I have no bias, although my wife's name is Susan.)

    -------------------------------------------
    Milton Goldsamt
    Consulting Research Psychologist & Survey Statistician
    -------------------------------------------








  • 38.  RE:Is SAS on the decline?

    Posted 01-30-2014 11:44
    I think someone asked if there was a comparison of the various programs anywhere. Wikipedia actually has a good comparison of them, with several tables/grids: https://en.wikipedia.org/wiki/Comparison_of_statistical_packages

    There's a separate section comparing availability of ANOVA procedures, regression procedures, graphics, etc. It looks like the most complete software packages are JMP, SAS, Stata, SPSS, MATLAB, NCSS, GenStat, and R.

    Personally, I've never heard of GenStat before.


    -------------------------------------------
    Gabriel Farkas
    -------------------------------------------








  • 39.  RE:Is SAS on the decline?

    Posted 01-30-2014 09:41
    Hoo-boy.  Has anyone been converted one way or the other yet?  Nevertheless, I can't resist chiming in.

    If you want to learn how to use R well, I suggest you read _Practical Common Lisp_ by Peter Seibel.  There's an html version available on the web for free.

    I'm only half joking.  The book doesn't mention R.  However, R is inspired by and patterned after Lisp (Scheme, specifically), and is essentially a distant poor cousin in the Lisp family, lacking some Lisp features but being vectorized and having an awful lot of analytical tools.  If you want to write code that's concise, manageable, and often works right on the first try, I recommend you leverage the Lisp elements of R, and to that it helps to see how Lisp programmers go about it.  In fact, learning Lisp could change your perspective on programming forever, regardless of language.

    Done well, I maintain that R is more productive at accomplishing complex tasks, while SAS is more productive at simpler tasks.  For instance, if you're pulling data from a database, chances are you're only a few merges and subsets from an analysis-ready data set.  SAS will get you there quickly.  If you have farther to go and the route is circuitous, R's features as a full-featured programming language (similar to Lisp or Python) can make your problem manageable.  This entails factoring the problem into a number of simpler problems, until you have a number of operators that each do one thing well.  And are easy to test.  Five layers of functions calling functions would not be unheard of.  I think 5 layers of SAS macros calling macros would make me pull my hair out, but this approach is positively encouraged in R.  So you can see it's a very different orientation.

    Also, with 5 layers of functions, if I factor the problem appropriately I can test all the functions, and then have high confidence the result is right.  Of course I can't give a 100% guarantee, and my own tests don't carry regulatory or QA clout, but they improve my own code.  It means a lot to feel personally I can stand behind my work.  If I had to solve a complex problem with SAS, I would lose confidence in the quality of the code, and would advise my boss or client that I couldn't offer the degree of confidence in the result that I would have with R.

    In fact, R is much more similar to Matlab and Python than to SAS.  As a programming language, you need to use it consistently to maintain your skill.  If you don't see complex problems coming your way, and are not interested in this sort of programming, R may not be the best choice for you.  On the other hand, no one can take R away from you, so any learning would not be totally wasted.  (Whether your employer will allow you to use it when you want to is another story.)

    Another selling point of R for me is reproducible (or dynamic) reporting, in which one mixes both narrative and statistical code in a source document.  When the report is "compiled" the code executes and the results are embedded.  No copying and pasting, no translation errors, and you always know what code generated what result.  Pharma companies like mine (Novartis) have elaborate infrastructure to accomplish this, but reproducible reporting and version control gives you essentially the same traceability and archiving capability, for free on your laptop, and in fact it's easier to use (than the elaborate infrastructure).  For R we have an embarrassment of riches:  Sweave, odfweave, Emacs Org mode, and knitr.  If you don't have any mechanism for archiving work and maintaining traceability of data and code to results, you should stop what you're doing and think about it now.  There's nothing R-specific about these tools.  I'm sure SAS, if they chose to, could implement something equivalent in a very short time.  But they haven't.  Or they have and it's so expensive I haven't heard of anyone using it.  I know there was a "SASweave" project that started but died out.  So for whatever reason, if you want reproducible reporting, R is the way to go for now.

    -Jim Garrett

    -------------------------------------------
    James Garrett
    Sr. Assoc. Dir. of Biostatistics
    Novartis
    -------------------------------------------








  • 40.  RE:Is SAS on the decline?

    Posted 01-30-2014 10:32
    Hello Group,

    I have been following this SAS etc discussion.  For someone looking for an affordable stand alone software with all of the statistical routines one could possible use, friendly RTF output, good documentation, excellent support, menu driven (vs R) I suggest taking a look at NCSS9 [NCSS.com].  Without a license you can buy a package for $329.00.  I cna't think of a better bargin.  I've used NCSS for over 25 years in my biostatistical practice.  Disclosure   I am not affiliated with NCSS just a user.  There is also a free trial period. 

    -------------------------------------------
    John Bartko
    Consulting Biostatistician
    -------------------------------------------








  • 41.  RE:Is SAS on the decline?

    Posted 01-30-2014 14:52
    I have thought about learning Lisp.

    If R is based on Scheme, should we not use a book about Scheme? Or is "Practical Common Lisp" really Scheme?

    Peter

    -------------------------------------------
    Peter Flom
    -------------------------------------------








  • 42.  RE:Is SAS on the decline?

    Posted 01-30-2014 14:57
    It's just that _Practical Common Lisp_ is a very accessible book that deals with realistic programming problems, and there is sufficient overlap between Common Lisp and Scheme.  To shape your thinking about programming, it's sufficient to learn Common Lisp.  There's no harm in learning Scheme as well, though.

    -Jim

    -------------------------------------------
    James Garrett
    Sr. Assoc. Dir. of Biostatistics
    Novartis
    -------------------------------------------