Andrew Cooper wrote:
Hello again S-news-ers,
After reading articles, help files, s-news and R archives, and, of
course V&R, I'm starting to understand regression with an ordinal
response variable. But I have a question (which may demonstrate that I
really don't understand regression with an ordinal response variable)...
I have an ordinal variable with 5 levels (e.g., high, medium-high,
medium, medium-low, and low) and 2 independent continuous variables.
Proportional odds logistic regression (e.g., polr() or lrm()) assumes
that the effect of each independent variable is constant across the 5
levels of the ordinal dependent variable. Which, as I understand it, is
at the heart of the proportional odds assumption.
Multinomial logit models (e.g., multinom()) allows the effect of the
independent variables to differ across the levels of the categorical
dependent variable, but does not use the information that our dependent
variable is an ordered response.
Is there something out there in S-plus that I could use when the
proportional odds assumption is violated that has the flexibility of the
multinomial logit (with different effects sizes by level) but still
accounts for the fact that I've got an ordered dependent variable? I
feel like I must be missing or not understanding something.
Thank you for your help!
Cheers,
Andy
A chapter in my book Regression Modeling Strategies has a detailed case
study of a continuation ratio ordinal logistic model with departures
from the equal slopes assumption.
Frank Harrell
******************************************************************
"What if the Hokey Pokey is all it really is about?" - Jimmy Buffett
Andrew B. Cooper, Ph.D.
Department of Natural Resources
Nesmith Hall, Rm 208A
University of New Hampshire
Durham, NH 03824
andrew.cooper@unh.edu
603.862.4254, 603.862.4976 (FAX)
http://www.unh.edu/natural-resources/fac-cooper.html
******************************************************************
--------------------------------------------------------------------
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
--
Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
|