[R] polr for ordered multinomial response without additional variables

Stats Wolf stats.wolf at gmail.com
Wed Feb 3 06:34:51 CET 2010


Dea all,

Let's suppose I am studying a questionnaire survey and one of the
questions has three ordered categorical responses (say, A, B and C).
Eg

result<-ordered(c(rep("A",12),rep("B",37),rep("C",6)))

Assume the respondents are not grouped. The differences between the
subsequent levels can be, I think, modeled with polr:

fit<-polr(result~1)
summary(fit)

Could you please explain me what do the coefficients from the above
summary mean? And how I could use the t-test (eg, what are degrees of
freedom, for example? What is the exact hypothesis being tested
here?)?

Thank you,
Stats Wolf



More information about the R-help mailing list