Max a écrit : > Prof Brian Ripley explained : >> On Mon, 26 Nov 2007, Max wrote: >> >>> Hi everyone, I'm trying to understand some R output here for ordinal >>> regression. I have some integer data called "A" split up into 3 ordinal >>> categories, top, middle and bottom, T, M and B respectively. >>> >>> I have to explain this output to people who have a very poor idea about >>> statistics and just need to make sure I know what I'm talking about >>> first. >>> >>> Here's the output: >>> >>> Call: >>> polr(formula = Factor ~ A, data = a, Hess = TRUE, method = "logistic") >>> >>> Coefficients: >>> Value Std. Error t value >>> A -0.1259028 0.04758539 -2.645829 >>> >>> Intercepts: >>> Value Std. Error t value >>> B|M -2.5872 0.5596 -4.6232 >>> M|T 0.3044 0.4864 0.6258 >>> >>> Residual Deviance: 204.8798 >>> AIC: 210.8798 >>> >>> I really am not sure what the intercepts mean at all. However, my >>> understanding of the coefficient of A is that as the category >>> increases, A decreases? If I have an A value of 10, how to I figure out >>> the estimated probability that this score is in one of the three >>> categories? >> Use predict(): see the book polr supports for examples (and the theory). > > I appreciate the reply, but have difficulty understanding what you mean > by "the book polr supports"? :-? > > The manuals in R don't reference the polr() command, nor do they write > about ordinal regression in R. (from what I can tell) The documentation > of the polr() doesn't explain the output or the theory... I've done web > searches on polr() and the MASS library and have found little of direct > help to my question.
Brian Ripley probably means "Modern Applied Statistics with S", W Venables and B. Ripley (4th edn), Springer, 2002. I'd also add "Categorical Data Analysis", Alan Agresti, Wiley (2000). HTH Emmanuel Charpentier ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.