[R] Regression models for ordinal responses ??

John Fox jfox at mcmaster.ca
Fri May 3 13:43:40 CEST 2002

At 09:50 AM 5/3/2002 +0200, you wrote:
>Hello list,
>Is there any mean to fit models for ordinal response other than multinomial
>polytomous ("multinom" from nnet ) and cumulative logit ("polr" from MASS)?
>I am particularly interested in continuation-ratio model and
>adjacent-category logit model. It is for the sake of epidemiology in
>wild-living populations!

Dear Emmaneulle,

Continuation-ratio models are just a collection of binomial (or binary) 
logit models, each of which can be fit with glm. So, for example, if you 
have categories A, B, and C, you'd fit a model in which the response is A 
vs. (B or C), and another in which the response is B vs. C. In the second 
model, the response if NA for those in category A. The models are 
independent, so, for example, you can add corresponding log-likelihood 
ratio statistics and df.

I believe that the adjacent-categories logit model is equivalent to the 
multinomial logit model. To get the coefficients for the log-odds of 
category i vs. i + 1 [i.e., log(pi_i/pi_{i+1})], subtract the multinomial 
logit coefficients for category i + 1 (vs. the baseline) from the 
coefficients for category i. (You should check that this is right -- I 
might not remember it correctly.)

I hope that this helps,
John Fox
Department of Sociology
McMaster University
Hamilton, Ontario, Canada L8S 4M4
email: jfox at mcmaster.ca
phone: 905-525-9140x23604
web: www.socsci.mcmaster.ca/jfox

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list