Tony should be statistically literate ;) to follow this through:
http://www.citeulike.org/user/ctacmo/article/574999. They ended up
concluding that every node of the tree consumes about three degrees of
freedom rather than one. Yes, trying to figure out where to stop in
growing those trees, and figuring out how to quantify the type I error
and do the inference is the damn difficult thing with those overly
flexible techniques.
On 12/16/08, Lachenbruch, Peter <[email protected]> wrote:
> I have also seen some studies (sorry I can't recall the authors) that suggest
> that CART over-fits models and provides more variables than are needed.
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
