I think what you're saying is that perhaps the response and the ordered
factor are related in a non-monotone way. That might be the case, but I
don't think it's a strong evidence. I'd guess it's due to the greedy nature
of the search. Besides, trees, being a nonparametric procedure, can still
pick out the non-monotone relationship, if it's really there, by having more
splits.
Cheers,
Andy
> -----Original Message-----
> From: D. Mckenzie [mailto:dmck@u.washington.edu]
> Sent: Thursday, January 30, 2003 1:27 PM
> To: John Fox
> Cc: Peter Flom; s-news@lists.biostat.wustl.edu
> Subject: Re: [S] Tree with ordinal independent variable
>
>
> Maybe I'm missing something (a common occurrence), but doesn't this
> solution beg the question that Peter's example answered: his
> tree model
> indicates that, with respect to the response variable, the
> factors are not
> ordered.
>
> __________________________________________________
>
> Don McKenzie, Fire and Landscape Ecologist
> Fire and Environmental Research Applications (FERA)
> Seattle Forestry Sciences Lab, USDA Forest Service
>
> donaldmckenzie@fs.fed.us
> dmck@u.washington.edu
> (206)732-7824; fax (206)732-7801
> __________________________________________________
>
>
> On Thu, 30 Jan 2003, John Fox wrote:
>
> > Dear Peter,
> >
> > There's no reason not to use numbers to represent the
> factor -- any numbers
> > in the correct order will do, such as integers 1 through 8.
> Splits will be
> > between adjacent numbers and hence will preserve the order
> of the factor.
> >
> > I hope that this helps,
> > John
> >
> > At 10:57 AM 1/30/2003 -0500, you wrote:
> > >Hello
> > >
> > >I am running S-Plus 6 on a Windows machine.
> > >
> > >I am trying to fit a classification tree with independent variables
> > >that are ordinal. However the result does not take the
> ordinalaity of
> > >the variables into account. e.g., one variable has levels
> A through H.
> > >I have made this an ordinal factor in the data window.
> But the tree
> > >gives splits such as
> > >
> > >AC BDEFGH
> > >
> > >I had thought of perhaps numericizing the scale, but this
> doesn't seem
> > >ideal to me. Is there a better solution?
> > >
> > >Thanks in advance
> > >
> > >Peter
> > >
> > >
> > >
> > >Peter L. Flom, PhD
> > >Assistant Director, Statistics and Data Analysis Core
> > >Center for Drug Use and HIV Research
> > >National Development and Research Institutes
> > >71 W. 23rd St
> > >New York, NY 10010
> > >(212) 845-4485 (voice)
> > >(917) 438-0894 (fax)
> > >
> >
> >--------------------------------------------------------------------
> > >This message was distributed by s-news@lists.biostat.wustl.edu. To
> > >unsubscribe send e-mail to
> s-news-request@lists.biostat.wustl.edu with
> > >the BODY of the message: unsubscribe s-news
> >
> > -----------------------------------------------------
> > John Fox
> > Department of Sociology
> > McMaster University
> > Hamilton, Ontario, Canada L8S 4M4
> > email: jfox@mcmaster.ca
> > phone: 905-525-9140x23604
> > web: www.socsci.mcmaster.ca/jfox
> > -----------------------------------------------------
> >
> > --------------------------------------------------------------------
> > This message was distributed by s-news@lists.biostat.wustl.edu. To
> > unsubscribe send e-mail to
> s-news-request@lists.biostat.wustl.edu with
> > the BODY of the message: unsubscribe s-news
> >
>
> --------------------------------------------------------------------
> This message was distributed by s-news@lists.biostat.wustl.edu. To
> unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with
> the BODY of the message: unsubscribe s-news
>
------------------------------------------------------------------------------
Notice: This e-mail message, together with any attachments, contains
information of Merck & Co., Inc. (Whitehouse Station, New Jersey, USA) that may
be confidential, proprietary copyrighted and/or legally privileged, and is
intended solely for the use of the individual or entity named on this message.
If you are not the intended recipient, and have received this message in error,
please immediately return this by e-mail and then delete it.
==============================================================================
|