s-news
[Top] [All Lists]

Re: Insightful - R cooperation

To: Terry Therneau <therneau@mayo.edu>
Subject: Re: Insightful - R cooperation
From: Frank E Harrell Jr <f.harrell@vanderbilt.edu>
Date: Fri, 26 Oct 2007 13:36:33 -0500
Cc: s-news@lists.biostat.wustl.edu
In-reply-to: <200710261506.l9QF6LF23387@hsrnfs-101.mayo.edu>
References: <200710261506.l9QF6LF23387@hsrnfs-101.mayo.edu>
User-agent: Thunderbird 1.5.0.14pre (X11/20071023)
Terry Therneau wrote:
Hi - Seeing Insightful sponsor R events, open up access to R within their
system, I'm wondering if this entails any risk to the development of R
itself. I assume that R's core development team does what's best for R, but
couldn't Insightful's influence prove to be a distraction at best? From an R
user standpoint, why spend energy aligning both systems, if such a thing is
actually being contemplated or even implemented?

I've seen this "abandon the old folks" attitude in some of the R community. Thankfully not very much of it because it really bothers me. I think that the stance is both arrogant and short sighted. (In fairness to the poster of the query above, I think that they are asking an honest question.) I've used S since the late 1980s -- I even have an AT&T source code license for it saved away in my nostalgia drawer. The switch to Splus was a big advance, they made substantial improvements to the code and have continued to do so. (If nothing else, cleaning up and expanding the help files; a thankless task that is none the less of great utility to a user). I've worked directly with the company since the early 90s, and they have consistently made my life easier wrt support and development of the survival code. There is room for both commercial and freeware entities in the market. Some places really do want paid support and a less agressive release cycle. Our group for instance often has a new release of SAS or Splus in house for about a year before making it the default. Predictability (knowing how it works) is more important than cutting edge. As one of my colleages put it when interviewing a candidate "This is not school: a B+ is not good enough. When working with researchers we have to get it all the answers right". This pressure can be even higher in pharma. That said, yes there is some effort in maintaining compatability. There are not a lot of differences between the R and Splus dialects of the S language, but a few are there. It is not difficult to write code that works with both packages, the survival package being an example. It's currently about 30 thousand lines, and has about 30 is.R() constructs (of which almost all involve oldClass vs class). Having the package in both systems is, to my mind, well worth that small level of effort. You can of course write R or Splus code that does not port well by seeking out the differences. But individual authors need to make up their own minds, and I suspect that Insightful will bear most of the burden of importing an R package. Terry Therneau

Terry and I have a friendly ongoing argument about some of these points. I feel the slow release cycle stifles innovation and inclusion of new features and bug corrections. A chaotic release cycle has always worked well for me.

Concerning compatibiliy, I spent hundreds of hours making the R Hmisc and Design packages compatible with S-Plus but now find that in the latest releases of S-Plus, which are supposed to be more R-compatible, that my packages are not included in the list of R packages that have been incorporated into S-Plus and hence the support for my packages has in some ways actually lessened instead of being enhanced.

Concerning S-Plus vs. R in terms of functionality and speed, there are very significant differences in the two.

Just my $.02 worth.

Frank

--
Frank E Harrell Jr   Professor and Chair           School of Medicine
                     Department of Biostatistics   Vanderbilt University

<Prev in Thread] Current Thread [Next in Thread>