[R] Likelihood ratio test in porl (MASS)

[Fox, John]

Dear Faradj Koliev,

There is an anova() method for "polr" objects that computes LR chisquare tests for nested models, so a short answer to your question is anova(Full, Restricted).

The question, however, seems to reflect some misunderstandings. First aov() fits linear analysis-of-variance models, which assume normally distributed errors. These are different from the ordinal regression models, such as the proportional-odds model, fit by polr(). For the former, F-tests *are* LR tests; for the latter, F-tests aren't appropriate.

I hope this helps,
 John

-----------------------------
John Fox, Professor
McMaster University
Hamilton, Ontario
Canada L8S 4M4
Web: socserv.mcmaster.ca/jfox




> -----Original Message-----
> From: R-help [mailto:r-help-bounces at r-project.org] On Behalf Of Faradj Koliev
> Sent: July 27, 2016 4:50 AM
> To: r-help at r-project.org
> Subject: [R] Likelihood ratio test in porl (MASS)
> 
> Dear all,
> 
> A quick question: Let’s say I have a full and a restricted model that looks
> something like this:
> 
> Full<- polr(Y ~ X1+X2+X3+X4, data=data, Hess = TRUE,  method="logistic”) #
> ordered logistic regression
> 
> Restricted<- polr(Y ~ X1+X2+X3, data=data, Hess = TRUE,  method="logistic”) #
> ordered logistic regression
> 
> I wanted to conduct the F-test (using aov command) in order to determine
> whether the information from the X4 variable statistically improves our
> understanding of Y.
> However, I’ve been told that the likelihood ratio test is a better alternative. So,
> I would like to conduct the LR test. In rms package this is easy -- lrest(Full,
> Restricted) — I’m just curious how to perform the same using polr. Thanks!
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.