Message-ID: <55e6eeab-0b26-4f4d-816f-787f29c1f41c@gmail.com>
Date: 2023-12-07T23:34:18Z
From: Ben Bolker
Subject: how to cite github resources'
In-Reply-To: <CAFSSRFL_d=0mXEtvA4FA_PqztGt+sGpOECTCGuYOGhzgJ2wYog@mail.gmail.com>
[please keep r-sig-mixed at r-project.org in the Cc: list]
On 2023-12-07 4:02 p.m., Antonio HERNANDEZ-MATIAS wrote:
> Many thanks for your message, I just need to justify that :
> With recent versions of |lme4|, goodness-of-fit (deviance) can be
> compared between |(g)lmer|?and |(g)lm|?models, although |anova()|?must
> be called with the mixed (|(g)lmer|) model listed first
Honestly, I don't think this is the kind of statement that needs a
citation from peer-reviewed literature or a book; instead, it's a
statement about the behaviour of software. The archaeology is a little
bit difficult, but I think the relevant information is from the [lme4
NEWS file](https://cran.r-project.org/web/packages/lme4/news.html),
which says for version 1.1-8 (2015-06-22) that:
* deviance() now returns the deviance, rather than half the negative
log-likelihood, for GLMMs fitted with Laplace (the behaviour for LMMs
and GLMMs fitted with nAGQ>1 has not changed)
This is consistent with the GLMM FAQ "git blame", which says that the
comment about "recent versions of lme4" was added on 2016-12-17 (at
which time version 1.1-8 would have been "recent")
Based on more archaeology, this change was made here, originally on
2014-07-02 (but the change wasn't in the CRAN version until the 1.1-8
release date mentioned above):
https://github.com/lme4/lme4/commit/5ac6f57f608e425765703b8e83eef7b513b599a6
As far as the requirement for listing the glmer() model first, that's
a quirk of the methods dispatch (more ugly detail is available if you
want it); IIRC listing the glmer() model other than first will throw an
error, rather than giving an incorrect answer.
If your supervisor/reviewers/whoever need more than a pointer to the
GLMM FAQ/a pointer to this e-mail in the list archives/a "pers. comm."
citation from me, you might be out of luck.
As long as we've gone this far down the rabbit hole, I'll also
emphasize the very next sentence in the GLMM FAQ, which points out a
potential *statistical* issue with such comparisons:
> Keep in mind that LRT-based null hypothesis tests are conservative
when the null value (such as ?2=0) is on the boundary of the feasible
space (Self and Liang 1987; Stram and Lee 1994; Goldman and Whelan
2000); in the simplest case (single random effect variance), the p-value
is approximately twice as large as it should be (Pinheiro and Bates 2000).
>
> El vie, 1 dic 2023 a las 22:48, Ben Bolker (<bbolker at gmail.com
> <mailto:bbolker at gmail.com>>) escribi?:
>
> ? ?If you really need a literature citation, you could specify exactly
> which piece of information from the GLMM FAQ you want to reference (the
> section has a lot of stuff in it) and someone might be able to give you
> a book or article reference that supports that information ...
>
> On 2023-11-30 10:06 a.m., Jeff Newmiller via R-sig-mixed-models wrote:
> > Isn't this driven by the conventions of the publishing context?
> >
> > On November 30, 2023 2:09:59 AM PST, Toni Hernandez-Matias
> <ahmatias at gmail.com <mailto:ahmatias at gmail.com>> wrote:
> >> Dear all,
> >> I would like to cite this information:
> >>
> https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#testing-significance-of-random-effects <https://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#testing-significance-of-random-effects>
> >>
> >> Do you know which format should I use?
> >>
> >> Thanks in advance,
> >>
> >> Toni
> >>
> >>? ? ? [[alternative HTML version deleted]]
> >>
> >> _______________________________________________
> >> R-sig-mixed-models at r-project.org
> <mailto:R-sig-mixed-models at r-project.org> mailing list
> >> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> <https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models>
> >
>
> _______________________________________________
> R-sig-mixed-models at r-project.org
> <mailto:R-sig-mixed-models at r-project.org> mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> <https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models>
>