I have been looking for some good statistics package for Common Lisp, that should be both stable and efficient. There is lhstats, but it doesn't have much documentation and it's not currently maintained. So what is the preferred way for statistics programming in Common Lisp (I'm using SBCL)? I'm looking for a package with advanced statistical functions, such as SEM (structural equation modeling), canonical correlations, GLM (general linear model) etc., but also even very basic but stable and well-documented statistical packages would be useful. Calling R statistical functions from Common Lisp would also be one option. What do you recommend?
15
votes
Why does it have to be Common Lisp? I think you will find R not very hard to learn coming from a Lisp background, and that way you won't be trying to reinvent the wheel....
– Ari B. Friedman
Yes, I was actually thinking of R, it has very good statistical functions and and I know the the basics of R syntax. I was looking for a Common Lisp solution, because it's more suitable for general-purpose programming IMHO, but maybe it's good to use R for statistics anyway, through RCL or RCLG.
– nrz
I think, if you find a good Common Lisp solution, you shouldn't go for R, just because it's common to use it for statistics. There's also an interesting article from R creator on the topic of Lisp use for stats: stat.auckland.ac.nz/~ihaka/downloads/Compstat-2008.pdf But if you'll decide to use R, you might want to look at swankr (a SWANK backend for R): common-lisp.net/~crhodes/swankr
– Vsevolod Dyomkin
@VsevolodDyomkin: I have read that article before and mostly I agree with it. Especially the issues of lack of multi-threading and relative numerical inefficiency of R when compared to Common Lisp are big issues for intensive numerical calculations, and for that reason using R for intensive numerical calculations is not an attractive option. For numerically non-intensive statistics using R does fine.
– nrz
common-lisp-stat is developed, just slowly. Personal challenges with the family have slowed me down. But I'm happy to take questions at [email protected].
– user1643706
4 Answers
11
votes
6
votes
The GNU Scientific Library has statistics functions and is accessible from GNU Scientific Library for Lisp.
4
votes
I think I would go with lhstat, it may not be maintained anymore, but the same thing seems true for common-lisp-stat (last commit is 1 year old), and at least it is available through Quicklisp. Moreover, maybe those packages aren't maintained because they're good enough.
You can also try RCL a FFI binding of R which is also quickloadable.
1
votes
I only know of it by name, but look here http://incanter.org/
It calls itself "a Clojure-based, R-like platform for statistical computing and graphics." Where clojure is a variant of common lisp.
