Dan Bron has just posted something, but I'll send this anyway.
Do you know APL? The free-standing "regression shelf" of the
"ASL" initiative, ie the APL Statistics Library from the
mid-90s, developed by Alan Sykes implemented a GLIM approach;
it has several link and distribution functions, including
the LOGIT link and the NORMAL distribution.
Translation to J would not be trivial - there are 94 APL
functions - but it might be worth undertaking, unless other
statisticians have done a J logit. Failing that, I suppose
you could use monadic matrix inverse on suitably transformed
variables. Ewart Shaw and others might also wish to comment.
Also in APL, way old, is Keith Smilley's statpak. Statpak 8
is a regression package, and could presumably be applied to
transformed variables.
I have both APL workspaces available in the dyalog APL
dialect.
Mike
Cox, Ed wrote:
Does anyone have code for logistic regression they could share? Preferably a
stepwise version. I periodically try to get interest in J started here at
work, but without a stepwise logit this is unlikely. Thus I have found it hard
to justify the time to really learn J. (Note - I asked this back in 2003 in
the General discussion to no avail.)
Ed Cox
Principal DB Analyst
CardinalHealth - Clinical Knowledge Services
_________________________________________________
This message is for the designated recipient only and may contain privileged,
proprietary, or otherwise private information. If you have received it in
error, please notify the sender immediately and delete the original. Any other
use of the email by you is prohibited.
Espanol - Deutsch - Nederlands - Francais - Italiano - Norsk - Svenska -
Japanese - Danish - Portuguese: www.cardinalhealth.com/legal/email
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
