R is an Epic Fail? | /en/2010/04/r-is-an-epic-fail/

Guest *Tal Galili* @ 2010-04-15 15:12:00 originally posted:

Hello Yihui Xie,
Yesterday I also came across the post, and wrote me thoughts about the subject here:
http://www.r-statistics.com/2010/04/r-the-next-big-thing-and-statistics-in-the-cloud/

Best,
Tal

Guest *Larry (IEOR Tools)* @ 2010-04-15 17:08:31 originally posted:

I couldn't agree more Yihui. I've started using R full time in the last year and I've found it to be very user friendly, easy to install, and loaded with documentation, mailing list support, and even IRC chat support in real time! I'm not sure what more you need? I think R should be a best-in-class application in my opinion.

Guest *Dan Goldstein* @ 2010-04-15 18:03:20 originally posted:

Hi Yihui,

I love your new site.

best,
Dan

Guest *brent* @ 2010-04-15 19:13:11 originally posted:

I wouldn't worry about it. The whole concept of "next big thing" is flawed and "user-friendly" only makes sense if you presume a certain type of user.

For the sort of people who do data analysis a lot, SPSS was much more user-friendly in the early 1990s: command-line driven with text-only output. As far as I can tell, the new SPSS GUI uses menus to constuct the same scripts and then runs them through the same engine; the GUI just adds an extra step. SAS is still much easier to script than it is to menu-menu-menu-select over and over again. Matlab has done fine being primarily command-line (and no less cryptic than R).

The GUIs opened a new market for these guys, but crippled their products. Thankfully, R doesn't need to woo shareholders with promises that by becoming shiny and "easy to use" it will dominate the world. Hopefully that will prevent R from becoming SPSS v17 for Mac: a serious a clunker.

JMP however, is a wonder to behold. It's a real GUI-driven stats package, not a command-line one with drop-down menus and dialog boxes replacing the typing. Between R and JMP I'll never need anything else.

Guest *Cody L. Custis* @ 2010-04-15 23:37:32 originally posted:

I’d leave my comment about the post on the original Julia Group post, but AnnMaria has a nasty tendency to withhold dissenting comments.

The difference between R and the other packages mentioned is that R’s combination of flexibility and similarity to other programming languages give it a major advantage. The example R code that AnnMaria derides is not only familiar to R programmers (who can see that it is not functional), but also to Matlab programmers, Python programmers, and C++ programmers (who also see that it is not functional). There is a lot to be said for tapping into commonality.

As someone who programs in SAS, I’m reminded that SAS has had regular expression searching since version 6, so roughly 20 years. However, version 9 introduced PERL regular expressions, which allowed SAS programmers to lean on another community for support, and greatly increased use within SAS.

R has the flexibility to take advantage of whatever communities are already out there, between its open source and extensive package system. R itself is not the ‘next big thing.’ However, R will be relevant and improved by the next big thing, whereas closed programming languages (SAS, SPSS, etc.) will be made irrelevant and obsolete.

Guest *CrisisMaven* @ 2010-04-16 04:17:50 originally posted:

As you are also dealing with the visualisation of statistical data: You may want to check out my Visualisation References resource list, aspiring to be the most comprehensive on the net. If you miss anything that I might be able to find for you or if you yourself want to share a resource, please leave a comment.

Guest *Kevin Wright* @ 2010-04-17 00:13:04 originally posted:

While installing R the first time is quite simple on Windows (quite a nuisance on linux when I tried it), the R codebase is changing dramatically from one version to the next. These rapid changes require that old R scripts be cleaned up and made compatible with the latest version. I spend a non-trivial amount of time trying to keep my code compatible with the changes in R. Some of the changes over the years have included the introduction of namespaces (which hides functions), addition of arguments to functions ('match' gained a new argument a few years ago), introduction of S4 methods, an apparently random move of a function from one package into another (stats to tools, maybe?), and now dynamic help pages. Nearly every major version of R creates incompatibilities with my scripts, and in that sense is an "Epic Fail".