R for the Rest of Us
April 29, 2010 4:07 PM   Subscribe

R is powerful, but tricky. RKWard is an UI that make R user friendly or, at least, make it more similar to SPSS or Stata. Screenshots!
posted by - (25 comments total) 44 users marked this as a favorite
I love the name!

As someone trying to learn R right now, yes, it can be awkward indeed.
posted by barnacles at 4:19 PM on April 29, 2010

I've tried a number of R GUIs. None of them have really done much for me, because they all seem to constrain you to just the feature set they decided to provide in the interface. I'm going to go try RKWard, I'll let you know.
posted by Jimbob at 4:26 PM on April 29, 2010 [1 favorite]

Oh hang on. Getting RKWard to work in Windows is, well. awkward. You need to get KDE and PHP (er, what?) going first. A project for another day, it seems.
posted by Jimbob at 4:27 PM on April 29, 2010 [1 favorite]

I just came across Red-R...
posted by stratastar at 4:33 PM on April 29, 2010 [1 favorite]

I thought Rscript was making R user friendly.
posted by benzenedream at 4:33 PM on April 29, 2010

Does this make R the Dwarf Fortress of statistics software?
posted by flaterik at 4:33 PM on April 29, 2010 [3 favorites]

If you like open source math programs you may want to look into Sage and KNIME.
posted by yoyo_nyc at 4:50 PM on April 29, 2010

Ooooooooh. Pretty.
posted by The White Hat at 5:17 PM on April 29, 2010

I don't think R needs this. It doesn't encourage thinking.
posted by cromagnon at 5:18 PM on April 29, 2010 [1 favorite]

whoa. I missed that earlier post. Wish I'd seen all that reference material when I was first learning it. It can be a damned nuisance when you've done something before, forgotten how to do it, and then spend hours fruitlessly googling for help when the answer is so simple.
posted by moorooka at 5:38 PM on April 29, 2010

quick and dirty R

posted by k8t at 5:52 PM on April 29, 2010 [2 favorites]

Kate is kinda nice for R. VIM4Life; there are a couple of decent plugins. I'm told that the eclipse add on is nice because of all the eclipse code tools.
posted by a robot made out of meat at 6:29 PM on April 29, 2010

This is pretty un-user friendly in an attempt to make R more user friendly. Still giving it a shot though...
posted by k8t at 7:06 PM on April 29, 2010


If you don't grok R, you're using the wrong tool. Trying to make it user friendly ruins it. R isn't easy, because the problem set it is attacking is very hard.

If you can't grok R -- if it absolutely escapes you, the right answer isn't trying to find a user friendly version of it. The right answer is find another profession, because if you can't understand R, you can't survive in a job that demands the skill sets that understanding R requires.

There's no shame in this. We all have our gifts, our strengths and weaknesses. But if you can't understand statistics and representing such graphically, R can't help you.

You're not RKWard, you're INCmptent. Find a profession that matches your skills.

As to the hacker that came up with this bogosity? That's why I don't run hacks in production.
posted by eriko at 8:46 PM on April 29, 2010 [1 favorite]

eriko: SPSS has a rep for having an easier UI than R doesn't it?

A number of the people I work with use SPSS because they find it easier to understand. What would be wrong with getting that sort of interface set up for R?
posted by sien at 8:57 PM on April 29, 2010

Jeez Eriko. You must have a lot of free time.

Cuz I understand the stats but don't have time to learn the syntax. Menus make it faster.

Menus don't let incompetent people magically run complex analysis.
posted by k8t at 9:41 PM on April 29, 2010 [5 favorites]

In general, point and click interfaces make it too easy to get yourself into trouble thinking you're running something you're not, and make it FAR too easy to perform a string of actions, get a result you don't understand or that seems neat, and then have no idea how you actually did what you did when you come and ask me how to interpret it.

But eriko is still mostly wrong.

I don't like point and click interfaces, but a more user-friendly command line and command syntax would not "ruin" R.

R isn't not-easy because the "problem it's attacking is very hard." There are lots and lots of statistical packages that people get far more easily than R. Even command-line-oriented packages that are highly extensible, like Stata. R is not-easy because its general structure assumes that you are more or less familiar with programming, even though many perfectly smart people who are thoroughly knowledgeable about statistics are not. It's biased very, very far towards being an environment that programmers find easy to program in rather than an environment that people use to apply well-established techniques to existing data.

And pointing at people who "absolutely can't" grok R is, well, rather silly. The problem with R, such as it is, is not that there are people who literally could not under any circumstances whatsoever understand how to use it, the problem is people who see the learning curve and believe that the time required to learn R is too expensive for the benefit of using it.

Further, connecting people who don't grok R to people who don't understand statistics is wrong-headed. Understanding how to use R does not require any firm understanding of statistics. Understanding how to use or even program in any statistical package does not require any really firm understanding of statistics, and there are plenty of people who make their livings as SAS coders who couldn't tell you when to use an ordered logit and when to use a count model for a DV that's an additive index. It probably does require being comfortable with object-oriented programming environments, on the other hand. But it is possible to learn how to use statistical techniques perfectly well, and know which to use when and what the various problems with each are, even though you do not know and are vaguely uncomfortable with object-orientation.
posted by ROU_Xenophobe at 9:57 PM on April 29, 2010 [10 favorites]

If you don't grok R, you're using the wrong tool. Trying to make it user friendly ruins it. R isn't easy, because the problem set it is attacking is very hard.

Agreed, overall, but R is WEIRD. I say this as someone who has used it at times daily and at times monthly for about 3 years now. I still find it weird. Every time I think I've really grokkked its weirdness, the R community produces something even weirder, like the lattice graphics package. And that's not solely because R is dealing with difficult problems but rather also because it's a language that flourished for a long time in some sort of evolutionary academic cul-de-sac. It's kind of a scripting language like php or python but it's also kind of like lisp and then it's also sometimes just a command-line interface for producing charts. For people familiar with most programming and/or scripting languages, R still takes awhile to adjust to.
posted by treepour at 10:02 PM on April 29, 2010 [1 favorite]

If you don't grok R, you're using the wrong tool.

Depends what you mean by "using". If you have a well defined set of R routines with a few different settings, it's reasonable to wrap it in an interface.

R has great strengths (and, to be sure, infuriating weaknesses) resulting from the small army of data analysts who've contributed to it, and lots of people can benefit from that, even if they're scared of programming in it. There are plenty of apps out there which call canned scripts from some kind of front-end.

More problematic is when you're exposing routines to users who do not have the sophistication to interpret the output properly. That's something that the developers and testers always have to think about carefully. But it's certainly not a problem unique to R!

eriko may be one of Neal Stephenson's Morlocks. And that's fine. But there's nothing wrong with rationing out the Eloi a little Morlock power.
posted by 7-7 at 12:40 AM on April 30, 2010

Indeed. The problem with Graphical interface packages for R is that R has modular packages, and those packages need their own GUI interface.

Realistically, somebody should create a GUI standard so that it can be included into packages by the authors or users of the work and interpreted by multiple GUIs.

But instead of working on a standard for GUI's, each GUI for R usually just replicates the base functions and maybe some popular packages, reinventing the wheel.

That, and R documentation is often, obtuse, terse, incomplete, highly specialized with no examples or discussions of actual application, or otherwise difficult. SAS is crappy, it's difficult to even use arrays for goodness sake, but many things are simpler to figure out how to do in SAS because they are better documented, and the documentation discussions assumptions, use cases, post hoc tests etc.

I do like R, and this looks interesting, especially in a Linux environment. But R really needs to be better organized. And that's not even discussing the actual language's pecularities.
posted by gryftir at 2:42 AM on April 30, 2010

I don't like point and click interfaces, but a more user-friendly command line and command syntax would not "ruin" R.

The syntax is more user friendly than Perl, C, Java or JavaScript, as far as I'm concerned, so that's good enough for me.

But R really needs to be better organized.

Definately, especially in terms of the documentation. For a lot of functions and libraries, you've got the man-page style help file, that usually tell you very little about what the function actually does and how to do it. And you've got some full, complete R-based statistical text books. But very little inbetween. It needs to be fleshed out a bit.
posted by Jimbob at 3:21 AM on April 30, 2010

I will add one more voice to say eriko is way off. R is powerful because of its scope. Including the libraries, there are very few things R can't do. The problem is that it looks like software made by 1000 hands, which it is. Even the core functions use different conventions and styles and once you add in the libraries it is a really big mess. Many others have already mentioned the lack of good documentation. I am a stats guy and a programmer, but I am not an every-day stats guy. I find R frustrating because when I return to the problems every few weeks I have to often relearn the idiosyncrasies of R to get anything done. Sometimes a problem that should take minutes goes to hours because it can be hard to find good documentation and examples.

If R is to find acceptance outside of the everyday-stats professionals it is going to need to (at minimum) fix its documentation gap. It would also benefit from some syntax and function consistency. Finally it needs a workable and extensible system for GUI support.

Despite my frustrations with R I find myself being a cheerleader for it. That is why I am very hopeful that RKWard or one of the other GUI projects out there will solve some of the problems to make R available to those outside the everyday-stats world.
posted by Tallguy at 11:09 AM on April 30, 2010

I took the opportunity of trying Ubuntu 10.04 to install RKWard.

I guess it offers about the level of functionality of, say, JMP Student Edition, that is, suitable for undergraduates and patients recovering from surgery.

I mean, I couldn't find the button for generalized linear models. Not that I would ever need a button to launch a GLM, but still...
posted by Jimbob at 3:35 PM on May 2, 2010

Is it just me or is there no ANOVA here?
posted by k8t at 11:07 PM on May 2, 2010

I've just found out that Revolution R Enterprise (a version of R with speed / multiprocessor optimizations) is currently free under academic licence. And it's very, very nice.
posted by Jimbob at 3:10 AM on May 7, 2010

« Older Hire illegals, lose your property.   |   Blast Off! Up to the stars we go! Newer »

This thread has been archived and is closed to new comments