[R] Ordered logit with polr won't match SPSS output

Juliet Hannah juliet.hannah at gmail.com
Thu Sep 30 16:15:56 CEST 2010


I think the most common reason to see different parameter estimates
with ordinal regression is that programs set up
the model differently.

For example, check out

library(MASS)
?polr

We see polr uses:

logit P(Y <= k | x) = zeta_k - eta

and notes that other software packages may use the opposite sign for
eta. Also check out that the ordered variables have
the same order (reference category). I sometimes have messed that up.



On Mon, Sep 27, 2010 at 2:53 PM, Ben Hunter <bjameshunter at gmail.com> wrote:
> I am learning R via a textbook that performs analysis with SPSS and SAS. In
> trying to reproduce the results for an ordinal logit model, I get very
> similar point estimates for my cut-off points, but the parameters for the
> covariate q60 do not match. The estimate for q51 also matches. Is this
> because I need to change a base case for the ordered covariate q60? Can this
> be done in or is it always the first case?
>
> mod<-polr(as.ordered(q43j)~as.ordered(q60)+q51, method="logistic")
>
> Perhaps a book using R would be a better idea, but it's the content and
> price (free) of this book that I like.
>
> Thanks a lot,
>
> Ben
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at 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.
>



More information about the R-help mailing list