Is R overtaking SAS at universities and colleges?
in [SAS]
|
From: "Keintz, H. Mark" on 30 Jun 2008 15:52 Not so fast. Let's get back to the original question, which was about SAS at universities and colleges and somehow transmogrified to what is used by statistics students who are going to work for pharma. So far I've only seen opinions based on personal experience of R and some quasi-anecdotal attribution of R overtaking SAS in the colleges and universities, whose truth so far has not be supported with evidence. Here's a survey of admittedly self-selected respondents from a pool of over 200 business schools (which I agree has little or nothing to do with clinical statistics, sociology, demography, and even most economics) . Most of the respondents are grad students at these schools: The question posed was "What Statistical Program do you use most?" Package N Rank SAS 143 1 Stata 21 2 SPSS 2 S+ 3 R 2 Matlab 16 3 Gauss 1 Eviews 3 TSP 0 Excel (and addins) 6 Other 1 I believe the high usage for SAS in the above table is partly due to the belief that SAS is better at handling large datasets and character variables. And the people responding to this survey are likely to be using datasets with these characteristics. In other academic groups with whom I've worked over the last several years, the economists prefer Stata to SAS (until they need to deal with large datasets), and statistics students learn R, or Matlab, because understanding and tweaking the statistical algorithm is of primary importance. Databases? Who needs 'em? I haven't worked with clinical data users in a long time, so maybe there are lots more R users at my campus that I haven't run into. But SAS is not sitting still in the academic world, so I wouldn't abandon it just yet. For the last year, they've been promoting SODA (SAS On Demand for Academics). SODA is a scheme whereby the putatively user-friendly graphical-interface enterprise guide is installed on the user's desktop (minimizing the agony of SAS installation), and it is connected to a server run by SAS Institute. SAS is obviously hoping that if the EG is sufficiently user-friendly and flexible especially for the classroom environment, comparisons that identify SAS with non-intuitive SAS BASE and SAS/Graph language will be out of date. And of course this user interface paradigm is also well-targeted to the business-intelligence crowd, the customer-flavor-of-the-month at SAS for a few years now. Regards, Mark > -----Original Message----- > From: SAS(r) Discussion [mailto:SAS-L(a)LISTSERV.UGA.EDU] On Behalf Of > RolandRB > Sent: Monday, June 30, 2008 3:24 PM > To: SAS-L(a)LISTSERV.UGA.EDU > Subject: Re: Is R overtaking SAS at universities and colleges? > > On Jun 30, 8:48 pm, ohri2...(a)GMAIL.COM (ajay ohri) wrote: > > Hi List > > > > I do use R on and off using the R gui's which help me speeden a lot > of > > modeling . You can click the *tag R* onwww.decisionstats.comwhere I > have > > written on them.(tag cloud is on the right) > > > > Take a look at the R GUI's , if u are thinking of transitioning. > > > > learning R syntax is more painful due to sheer number of functions > (than the > > limited multi purpose procs available in SAS), but R online > documentation > > on the sites as well download mirrors make it easy to install > appropriately. > > > > as for reporting R graphics are definitely much superior, and I > recommend > > downloading the R GUI called rattle for anyone . Just install it , > and use > > it. The code gets auto generated so u learn basic R programming even > fast. > > > > Yes , the transition is quite painful but two weeks with the R gui > and > > search R web documentation can do a lot of work (merging/sorting big > > datasets very very fast is still SAS 's plus point) > > I can see your point. If sas is gone then data will still need > transforming and that will be huge amounts of data if lab data for > clinical trials. With no sas database then I guess it will be Oracle > or MySQL and the transformation done using SQL but I guess it can be > done. I am so used to using sas now that it will be hard but I've used > Basic and Cobol in the past and was able to do whatever I liked with > the data I was working on. I guess there is a way. > > I wonder what statisticians fresh out of college and now working for > pharmas think about this. Do you prefer R and get forced down the sas > path? Would you prefer the programmers to help you with R rather than > giving you solutions using sas? Do you think sas is a dead language > for the analysis of clinical trials?
From: BruceBrad on 30 Jun 2008 20:12 >But SAS is not sitting still in the academic world, so I wouldn't abandon it just yet. For the last year, they've been promoting SODA (SAS On Demand for Academics). SODA is a scheme whereby the putatively user-friendly graphical-interface enterprise guide is installed on the user's desktop (minimizing the agony of SAS installation), and it is connected to a server run by SAS Institute. SAS is obviously hoping that if the EG is sufficiently user-friendly and flexible especially for the classroom environment, comparisons that identify SAS with non-intuitive SAS BASE and SAS/Graph language will be out of date. SODA might be useful for teach statistics, but enterprise guide-type interfaces are not the way to do research. Code editing is still the best way to organise and document research. Among economists and related social sciences, Stata is really taking off because of its speedy incorporation of new statistical methods. I haven't used it, but R has the potential to do the same, particularly if Stata gets too greedy in its pricing. SAS is a slow dinosaur by comparison. To take two examples: - Proc Panel is still not yet in production, and the pre-production version doesn't handle large tables well (in my brief experience). I know it is in in production in v9.2 (which might work better), but we can't get that yet. I know you can do many panel estimation procedures in other procedures, but they don't have the ease of use of Stata's xtreg. - Replication methods are much more straightforward to use in Stata.
From: ohri2007 on 1 Jul 2008 01:17 part of the reason business school students like SAS is easy to learn and they dont have to study statistics too much to shape data.same is the reason for SPSS. write a few procs in SAS or click a few buttons in SPSS and the group project is complete, which is quite nifty in the hectic schedule of a business school student (i used to be one ...:) sas as a software is good still its the pricing, lack of innovation ,collaborative working that will hurt it compared to R.....and SODA seems nothing but a way to get people hooked on to SAS without realizing some things are better done in other softwares.. overselling in short term ? regards, ajay www.decisionstats.com On 7/1/08, Keintz, H. Mark <mkeintz(a)wharton.upenn.edu> wrote: > Not so fast. > > Let's get back to the original question, which was about SAS at universities > and colleges and somehow transmogrified to what is used by statistics > students who are going to work for pharma. > > So far I've only seen opinions based on personal experience of R and some > quasi-anecdotal attribution of R overtaking SAS in the colleges and > universities, whose truth so far has not be supported with evidence. > > > > Here's a survey of admittedly self-selected respondents from a pool of over > 200 business schools (which I agree has little or nothing to do with > clinical statistics, sociology, demography, and even most economics) . Most > of the respondents are grad students at these schools: > > The question posed was "What Statistical Program do you use most?" > > Package N Rank > SAS 143 1 > Stata 21 2 > SPSS 2 > S+ 3 > R 2 > Matlab 16 3 > Gauss 1 > Eviews 3 > TSP 0 > Excel (and addins) 6 > Other 1 > > I believe the high usage for SAS in the above table is partly due to the > belief that SAS is better at handling large datasets and character > variables. And the people responding to this survey are likely to be using > datasets with these characteristics. > > In other academic groups with whom I've worked over the last several years, > the economists prefer Stata to SAS (until they need to deal with large > datasets), and statistics students learn R, or Matlab, because understanding > and tweaking the statistical algorithm is of primary importance. Databases? > Who needs 'em? > > I haven't worked with clinical data users in a long time, so maybe there are > lots more R users at my campus that I haven't run into. > > > But SAS is not sitting still in the academic world, so I wouldn't abandon it > just yet. For the last year, they've been promoting SODA (SAS On Demand for > Academics). SODA is a scheme whereby the putatively user-friendly > graphical-interface enterprise guide is installed on the user's desktop > (minimizing the agony of SAS installation), and it is connected to a server > run by SAS Institute. SAS is obviously hoping that if the EG is > sufficiently user-friendly and flexible especially for the classroom > environment, comparisons that identify SAS with non-intuitive SAS BASE and > SAS/Graph language will be out of date. > > And of course this user interface paradigm is also well-targeted to the > business-intelligence crowd, the customer-flavor-of-the-month at SAS for a > few years now. > > > Regards, > Mark > >> -----Original Message----- >> From: SAS(r) Discussion [mailto:SAS-L(a)LISTSERV.UGA.EDU] On Behalf Of >> RolandRB >> Sent: Monday, June 30, 2008 3:24 PM >> To: SAS-L(a)LISTSERV.UGA.EDU >> Subject: Re: Is R overtaking SAS at universities and colleges? >> >> On Jun 30, 8:48 pm, ohri2...(a)GMAIL.COM (ajay ohri) wrote: >> > Hi List >> > >> > I do use R on and off using the R gui's which help me speeden a lot >> of >> > modeling . You can click the *tag R* onwww.decisionstats.comwhere I >> have >> > written on them.(tag cloud is on the right) >> > >> > Take a look at the R GUI's , if u are thinking of transitioning. >> > >> > learning R syntax is more painful due to sheer number of functions >> (than the >> > limited multi purpose procs available in SAS), but R online >> documentation >> > on the sites as well download mirrors make it easy to install >> appropriately. >> > >> > as for reporting R graphics are definitely much superior, and I >> recommend >> > downloading the R GUI called rattle for anyone . Just install it , >> and use >> > it. The code gets auto generated so u learn basic R programming even >> fast. >> > >> > Yes , the transition is quite painful but two weeks with the R gui >> and >> > search R web documentation can do a lot of work (merging/sorting big >> > datasets very very fast is still SAS 's plus point) >> >> I can see your point. If sas is gone then data will still need >> transforming and that will be huge amounts of data if lab data for >> clinical trials. With no sas database then I guess it will be Oracle >> or MySQL and the transformation done using SQL but I guess it can be >> done. I am so used to using sas now that it will be hard but I've used >> Basic and Cobol in the past and was able to do whatever I liked with >> the data I was working on. I guess there is a way. >> >> I wonder what statisticians fresh out of college and now working for >> pharmas think about this. Do you prefer R and get forced down the sas >> path? Would you prefer the programmers to help you with R rather than >> giving you solutions using sas? Do you think sas is a dead language >> for the analysis of clinical trials? >
From: RolandRB on 1 Jul 2008 12:29 On Jun 30, 3:36 pm, peterflomconsult...(a)mindspring.com (Peter Flom) wrote: > RolandRB <rolandbe...(a)HOTMAIL.COM> wtote > > >An email I received in reply to this is that R has overtaken SAS in > >colleges and universities like the race has already been run and won. > >I am wondering what knock-on effect this will have for the current in- > >house reporting systems that the pharmaceutical companies use. It is > >the statisticians who drive the analysis and if they are becoming "R > >savvy" and have useful skills with R then this may not sit well with > >the current sas reporting systems. If R is up to the job then I can > >see it replacing sas some time in the future. I've dabbled with R from > >time to time but I tend to learn from example and can't make the jump > >in my mind to using R for the sort of reports you can link to in the > >pps below rather than using sas macros. I wonder if it can be done. > >Can anyone advise? If they had R code then so much the better. > > >http://www.datasavantconsulting.com/roland/unistats.pps > > Roland > > I know almost nothing about pharma reporting, but I took a quick look at the powerpoint, and I bet that something > like that could be done in R using sweave. I've not *done* anything like that, because I don't have a standard format > that I need all the time, so mastering sweave hasn't been high on my list.. > > But that's where I'd say you should look, if you want to see what R can do along these lines > > And, after investigating that, you might want to ask on R-help.... but only after investigating. They have a posting guide, and they get upset when people don't follow it. > > Peter > > Peter L. Flom, PhD > Statistical Consultant > www DOT peterflom DOT com And now I find there's SASweave which can run both sas and R so if R wasn't up to the heavy tasks like lab data then use SAS for that instead. http://www.cs.uiowa.edu/~rlenth/SASweave/
From: "Keintz, H. Mark" on 1 Jul 2008 15:08
Ajay said: > -----Original Message----- > From: ohri2007(a)gmail.com [mailto:ohri2007(a)gmail.com] > Sent: Tuesday, July 01, 2008 1:17 AM > To: Keintz, H. Mark > Cc: SAS-L(a)listserv.uga.edu > Subject: Re: Is R overtaking SAS at universities and colleges? > > part of the reason business school students like SAS is easy to learn > and they dont have to study statistics too much to shape data.same is > the reason for SPSS. > > write a few procs in SAS or click a few buttons in SPSS and the group > project is complete, which is quite nifty in the hectic schedule of a > business school student (i used to be one ...:) > > sas as a software is good still its the pricing, lack of innovation > ,collaborative working that will hurt it compared to R.....and SODA > seems nothing but a way to get people hooked on to SAS without > realizing some things are better done in other softwares.. > overselling in short term ? I agree that SODA is a way to gain adherents to SAS before they leave school, which is perfectly consistent with SAS's strategy of charging universities less than 10% of commercial prices. I also agree that more generally SAS long ago moved its primary development effort from statistical analysis to various enterprise goals, like CDISC and business intelligence. And I imagine that developing better graphics would also take priority over new statistical apps. I imagine SAS sees ORACLE and SAP as competitors more than Stata, R, or Matlab. And the ability to work with Excel is probably an equally strategic goal. But the original assertion/question was not whether SAS has objectives that align with those of choose-your-favorite-academic-discipline, but that R is overtaking SAS at universities and colleges. It may be true, but so far nobody has offered evidence. > > regards, > > ajay > www.decisionstats.com > > On 7/1/08, Keintz, H. Mark <mkeintz(a)wharton.upenn.edu> wrote: > > Not so fast. > > > > Let's get back to the original question, which was about SAS at > universities > > and colleges and somehow transmogrified to what is used by statistics > > students who are going to work for pharma. > > > > So far I've only seen opinions based on personal experience of R and > some > > quasi-anecdotal attribution of R overtaking SAS in the colleges and > > universities, whose truth so far has not be supported with evidence. > > > > > > > > Here's a survey of admittedly self-selected respondents from a pool > of over > > 200 business schools (which I agree has little or nothing to do with > > clinical statistics, sociology, demography, and even most economics) > . Most > > of the respondents are grad students at these schools: > > > > The question posed was "What Statistical Program do you use most?" > > > > Package N Rank > > SAS 143 1 > > Stata 21 2 > > SPSS 2 > > S+ 3 > > R 2 > > Matlab 16 3 > > Gauss 1 > > Eviews 3 > > TSP 0 > > Excel (and addins) 6 > > Other 1 > > > > I believe the high usage for SAS in the above table is partly due to > the > > belief that SAS is better at handling large datasets and character > > variables. And the people responding to this survey are likely to be > using > > datasets with these characteristics. > > > > In other academic groups with whom I've worked over the last several > years, > > the economists prefer Stata to SAS (until they need to deal with > large > > datasets), and statistics students learn R, or Matlab, because > understanding > > and tweaking the statistical algorithm is of primary importance. > Databases? > > Who needs 'em? > > > > I haven't worked with clinical data users in a long time, so maybe > there are > > lots more R users at my campus that I haven't run into. > > > > > > But SAS is not sitting still in the academic world, so I wouldn't > abandon it > > just yet. For the last year, they've been promoting SODA (SAS On > Demand for > > Academics). SODA is a scheme whereby the putatively user-friendly > > graphical-interface enterprise guide is installed on the user's > desktop > > (minimizing the agony of SAS installation), and it is connected to a > server > > run by SAS Institute. SAS is obviously hoping that if the EG is > > sufficiently user-friendly and flexible especially for the classroom > > environment, comparisons that identify SAS with non-intuitive SAS > BASE and > > SAS/Graph language will be out of date. > > > > And of course this user interface paradigm is also well-targeted to > the > > business-intelligence crowd, the customer-flavor-of-the-month at SAS > for a > > few years now. > > > > > > Regards, > > Mark > > > >> -----Original Message----- > >> From: SAS(r) Discussion [mailto:SAS-L(a)LISTSERV.UGA.EDU] On Behalf Of > >> RolandRB > >> Sent: Monday, June 30, 2008 3:24 PM > >> To: SAS-L(a)LISTSERV.UGA.EDU > >> Subject: Re: Is R overtaking SAS at universities and colleges? > >> > >> On Jun 30, 8:48 pm, ohri2...(a)GMAIL.COM (ajay ohri) wrote: > >> > Hi List > >> > > >> > I do use R on and off using the R gui's which help me speeden a > lot > >> of > >> > modeling . You can click the *tag R* onwww.decisionstats.comwhere > I > >> have > >> > written on them.(tag cloud is on the right) > >> > > >> > Take a look at the R GUI's , if u are thinking of transitioning. > >> > > >> > learning R syntax is more painful due to sheer number of functions > >> (than the > >> > limited multi purpose procs available in SAS), but R online > >> documentation > >> > on the sites as well download mirrors make it easy to install > >> appropriately. > >> > > >> > as for reporting R graphics are definitely much superior, and I > >> recommend > >> > downloading the R GUI called rattle for anyone . Just install it , > >> and use > >> > it. The code gets auto generated so u learn basic R programming > even > >> fast. > >> > > >> > Yes , the transition is quite painful but two weeks with the R gui > >> and > >> > search R web documentation can do a lot of work (merging/sorting > big > >> > datasets very very fast is still SAS 's plus point) > >> > >> I can see your point. If sas is gone then data will still need > >> transforming and that will be huge amounts of data if lab data for > >> clinical trials. With no sas database then I guess it will be Oracle > >> or MySQL and the transformation done using SQL but I guess it can be > >> done. I am so used to using sas now that it will be hard but I've > used > >> Basic and Cobol in the past and was able to do whatever I liked with > >> the data I was working on. I guess there is a way. > >> > >> I wonder what statisticians fresh out of college and now working for > >> pharmas think about this. Do you prefer R and get forced down the > sas > >> path? Would you prefer the programmers to help you with R rather > than > >> giving you solutions using sas? Do you think sas is a dead language > >> for the analysis of clinical trials? > > |