Free / open statistical software

March 24, 2005 10:21 PM

Help me find free and/or open-source statistical software.

I'm teaching the fresh-faced young grad students an intro quant course -- the one where you start with "Here's our friend, the probability!" and work your way up through basic multiple regression.

Naturally, some of them want software to run their numbers at home, and don't necessarily want to pay the moderate cost for one of the school's Stata licenses. I could point them at R, but the first time they have to query an object they'd either run away screaming or lynch me. I could point them at gretl, but it's very GUI-oriented (yeah, I know there's a console), and I don't want them to get into the bad habit of dragging and clicking instead of building a solid history of commands that they can refer to later. Especially I don't want to have to deal with them in two years when they can't remember how they arrived at a particular set of results.

So, to what should I point them? I don't want to give them a fright, but I also would prefer something that's more command-line oriented than gretl. Something like Stata6, but free, and I don't care if it won't do fancy-schmancy stuff.

I assume they'll be running Wintel machines.

posted by ROU_Xenophobe to computers & internet (11 comments total)

• OpenStat (Windows + x86 Linux)
• SalStat (python-based, multiple platforms)
• NIST DataPlot (Windows-only)
posted by AlexReynolds at 11:17 PM on March 24, 2005

Hmm...if R is too much, I've heard good things about the open-source Matlab clone Octave - but I can't vouch personally for it, nor for its dev/support status.

What about just getting them using a simple language like Python or even Perl? There are plenty of good stats and analysis packages for these languages; they're freely available for every platform; and realistically they're not much harder to learn than a specific application language while being much more generally useful.
posted by freebird at 11:22 PM on March 24, 2005

Thanks, but the odds of them learning a programming language approach the odds of me being made Pope. This is a very basic class primarily for people who haven't had anything more than Calculus 1, and maybe that 5+ years ago.

I'll look into the packages above, but I suspect that getting (some of) them to install programs with manual dependency chains will be an uphill struggle.
posted by ROU_Xenophobe at 12:16 AM on March 25, 2005

I'll look into the packages above, but I suspect that getting (some of) them to install programs with manual dependency chains will be an uphill struggle.

That's the pain of most open source software, unfortunately.
posted by AlexReynolds at 12:47 AM on March 25, 2005

ROU_Xenophobe. I feel for you, what about a bootable linux distribution on CD with a couple of the packages pre-installed? It looks like Knoppix is instantly bootable and I've heard good things about it. A customized CD with Matlab or such installed might be possible.

A bootable DVD would fit it for sure.
posted by substrate at 5:24 AM on March 25, 2005

how about this?
found here via here - i was thinking that really you just need a stats-aware calculator at this level.
if it's for a limited time, or they have $25, then this looks pretty good.
posted by andrew cooke at 5:30 AM on March 25, 2005

oh, sorry, missed your bit about wanting these people to have to program (which strikes me as batty, but it's your call...)
posted by andrew cooke at 6:23 AM on March 25, 2005

At this point we're trying to convince them that they can do basic regression, understand the output, and interpret other people's results, all in a relatively friendly and confidence-building way, and trying to get them to learn another set of whole new skills doesn't fit with that.

We'll smack them around again in other courses, and people who end up doing work of any complexity will pick up basic programming skills as they go.

I thought about a bootable cd, but unfortunately I don't think I can bet that their average machine at home is new enough to boot from cd.

Sigh. I suppose I can tell them to use gretl... I really like that it makes diagnostic plots so easy. I just wish it had a better command line. Or to suck it up and use R.
posted by ROU_Xenophobe at 7:15 AM on March 25, 2005

There's another list of apps from AlexR's OpenStat link. A few of the screenshots -suggest- that some of them might be a little friendlier, if GUI=friendly, in this case.
posted by bachelor#3 at 10:36 AM on March 25, 2005

...another shout for octave. Matlab is dead easy, and is ideal for academic-y folk who don't want to worry about the little things (like types for variables).

That's including those who don't know much math (I had a math-anxious bio prof who swore by it)
posted by leotrotsky at 10:45 AM on March 25, 2005

ROU_Xenophobe, there's a web interface to Octave here. Your students could have access without any resource requirements on their side.
posted by substrate at 2:28 PM on March 26, 2005