Java package for statistical analysis?

January 23, 2006 4:22 PM

Know of a good, free Java package for basic statistical analysis?

I'm working on a product that needs to do some basic statistical analysis of data. At the moment the specific task at hand is linear regression, but I wouldn't mind having one place to go for simple standard deviation and stuff like that, too. I could spend a bunch of time writing stuff myself, but I figure, why reinvent the wheel?

The product I work on is free and open source, and that's what I'm looking for in a stats API, too. Or at least free! Also I want something as targeted as possible -- I don't want to pull in gigabytes of code just so I can use a couple classes.

I've been poking around looking for stuff, but surprisingly I've found little so far (this, for instance, is not free). Can anyone help me out?

posted by gurple to Computers & Internet (8 answers total)

JRI will let you run R from inside Java calls. Both are open source and free.
posted by Rothko at 7:46 PM on January 23, 2006

It's not Java and it's not open source, but you can download a student version of SPSS that's fully functional for free on their website. I think it expires after 30 days or so, but it does work. And SPSS is relatively user friendly.
posted by mulkey at 8:02 PM on January 23, 2006

Here's a site with a list of free stats software. Look around this website, it has tons of links, and this one might interest you.
posted by dhruva at 11:05 PM on January 23, 2006

Perhaps I am misunderstanding the question, but I think gurple is looking for a stats API he can use *within* his Java program. This rules out actual stats software, even web-based.

I searched around a bit and all I came up with was the same link referenced in the original post. If it were me I would probably just write the stuff I needed from scratch, or shell out the dough for the WebCab stuff.
posted by sanitycheck at 1:09 AM on January 24, 2006

JRI will let you run R, a popular and free open source stat pack, using Java methods. JRI provides an "API", roughly speaking.
posted by Rothko at 7:10 AM on January 24, 2006

Response by poster: Hmm, yep, it's not looking good so far. I use R heavily, and in fact my current workaround will be to call out to R, but that's cumbersome and not very efficient. I might end up just writing some linear regression code. It seems unbelievable that someone hasn't done this already and made it public!

Thanks for all the help so far!
posted by gurple at 10:56 AM on January 24, 2006

Whoops, didn't mean to include your answer in my generalization, Rothko. But would a programmer be able to package R with their open source product? I'm not up on the rules of redistributing OSS.
posted by sanitycheck at 10:56 AM on January 24, 2006

Response by poster: In fact, some of the people involved in developing R suggested the same thing to me -- package up a minimal R installation with our product. So I _think_ there are no legal issues. But still, it's a lot of dependency for a little functionality.
posted by gurple at 11:05 AM on January 24, 2006