On Thu, May 20, 2010 at 10:41 AM, Jason Grout
<[email protected]> wrote:
> On 5/20/10 9:56 AM, David Kirkby wrote:
>>
>> I gave a talk last night at the London Open Solaris User Group (LOSUG)
>> with the title "Porting Sage open source mathematics software to
>> OpenSolaris". I've stuck a copy of the presentation at
>>
>>
>> http://boxen.math.washington.edu/home/kirkby/talks/Sage-LOSUG-19-5-2010--by-David-R-Kirkby.odp
>>
>> The talk generated quite a bit of interest - I lost count of the
>> number of questions. Someone emailed me today, to ask about
>> statisticians using Sage. I won't forward his message, since I don't
>> have his permission to do so, but the relevant bit is:
>>
>> ---------------------------------------------------
>> I had one specific question that I didn't think was of general
>> interest: I work in a statistics research unit and I had already
>> downloaded (your?) solaris-10 sparc build before the talk.
>>
>> I wanted to ask, do you think sage offers a statistician (as opposed
>> to a mathematician)?  Perhaps I should ask if you know of any
>> statisticians already using sage?
>> ----------------------------------------------------
>>
>> I'm sure there are people far better placed to answer that question than
>> me.
>>
>
> One obvious thing is that Sage includes R, and makes it possible to work
> with R via the notebook (though some of the integration is rough around the
> edges).
>
> Of course, if they just wanted to run R on Solaris, then
> http://cran.r-project.org/bin/solaris/ might be a better choice.  If they
> wanted to use the notebook to share information, collaborate, etc., Sage
> still might be useful.
>
> Jason
>

I've been doing some statistics with Sage lately.  Here are some
reasons why I think Sage is a very powerful tool for stats:

   * The Notebook -- obvious

   * Sage includes R, and has both a pexpect and C-library (via rpy2)
interface to R, which works.

   * As a language, Python is vastly superior to R.   Python has good
support for object oriented programming, a very wide selection of
existing programs and libraries, and supports threads for handling
realtime data.    I recently read a paper about massive contortions
somebody went through in trying to be build some system in R to model
and respond to realtime data -- this was really hard in R, since R
evidently doesn't have good support for threads.  But of course, R +
Sage (via rpy2) would make it easy to combine the modeling power of R
with the asynchronous capabilities of Python.

  * Cython: many useful statistical models aren't really that
complicated (especially compared to the algorithms number theorist and
algebraists and even symbolic calculus folks deal with) -- if you
understand what you're doing, you can code them up from scratch, in
anywhere from a few hours to a week.  The results can be very
efficient if you write them using Cython, either directly using C data
types or against numpy (using the support Dag added).        Many
companies that make very serious use of statistics write code for what
they do "from scratch", since the added flexibility and power make
this worthwhile.  Sage (via tight Cython integration, the Numpy
library, etc.) provides a great platform for doing so.

  * Numpy/Scipy: They have a lot of statistical functionality built
in.   Some of it is very efficient. It's constantly improving.

  * Sage has a surprising amount of native basic stats now, including basic
descriptive statistics for any Sage iterable, and also a new fairly complete
generalized hidden markov model library.   I would like to see far more stats
natively implemented in Sage.  Type "stats.[tab]".

  * Python rocks. :-)

William

-- 
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/sage-support
URL: http://www.sagemath.org

Reply via email to