F and Wald chi-square tests in mixed-effects models

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

F and Wald chi-square tests in mixed-effects models

Helios de Rosario
I have a doubt about the calculation of tests for fixed effects in
mixed-effects models.

I have read that, except in well-balanced designs, the F statistic that
is usually calculated for ANOVA tables may be far from being distributed
as an exact F distribution, and that's the reason why the anova method
on "mer" objects (calculated by lmer) do not calculate the denominator
df nor a p-value. --- See for instance Douglas Bates' long post on this
topic, in:
https://stat.ethz.ch/pipermail/r-help/2006-May/094765.html

However, Anova does calculate p-values from Wald chi-square tests for
fixed terms from "mer" objects (as well as from "lme" objects, from
lme). I suppose that the key to understand the logic for this is in Fox
& Weisberg's commentary in "An R Companion to Applied Regression" (2nd
edition, p. 272), where they say: "Likelihood ratio tests and F tests
require fitting more than one model to the data, while Wald tests do
not."

Unfortunately, that's too brief a commentary for me to understand why
and how the Wald test can overcome the deficiencies of F-tests in
mixed-effects models. The online appendix of "An R Companion..." about
mixed-effects models does not comment on hypothesis tests either.

I would appreciate if someone can give some clues or references to read
about this issue.

Thanks,
Helios

INSTITUTO DE BIOMECÁNICA DE VALENCIA
Universidad Politécnica de Valencia • Edificio 9C
Camino de Vera s/n • 46022 VALENCIA (ESPAÑA)
Tel. +34 96 387 91 60 • Fax +34 96 387 91 69
www.ibv.org

  Antes de imprimir este e-mail piense bien si es necesario hacerlo.
En cumplimiento de la Ley Orgánica 15/1999 reguladora de la Protección
de Datos de Carácter Personal, le informamos de que el presente mensaje
contiene información confidencial, siendo para uso exclusivo del
destinatario arriba indicado. En caso de no ser usted el destinatario
del mismo le informamos que su recepción no le autoriza a su divulgación
o reproducción por cualquier medio, debiendo destruirlo de inmediato,
rogándole lo notifique al remitente.

______________________________________________
[hidden email] 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.
Reply | Threaded
Open this post in threaded view
|

Re: F and Wald chi-square tests in mixed-effects models

bbolker
Helios de Rosario <helios.derosario <at> ibv.upv.es> writes:

>
> I have a doubt about the calculation of tests for fixed effects in
> mixed-effects models.
>
> I have read that, except in well-balanced designs, the F statistic that
> is usually calculated for ANOVA tables may be far from being distributed
> as an exact F distribution, and that's the reason why the anova method
> on "mer" objects (calculated by lmer) do not calculate the denominator
> df nor a p-value. --- See for instance Douglas Bates' long post on this
> topic, in:
> https://stat.ethz.ch/pipermail/r-help/2006-May/094765.html
>
> However, Anova does calculate p-values from Wald chi-square tests for
> fixed terms from "mer" objects (as well as from "lme" objects, from
> lme). I suppose that the key to understand the logic for this is in Fox
> & Weisberg's commentary in "An R Companion to Applied Regression" (2nd
> edition, p. 272), where they say: "Likelihood ratio tests and F tests
> require fitting more than one model to the data, while Wald tests do
> not."
>
> Unfortunately, that's too brief a commentary for me to understand why
> and how the Wald test can overcome the deficiencies of F-tests in
> mixed-effects models. The online appendix of "An R Companion..." about
> mixed-effects models does not comment on hypothesis tests either.
>
> I would appreciate if someone can give some clues or references to read
> about this issue.

  Can you please repost this to the r-sig-mixed-models list?  I think this
is an important point and may get lost in the noise here.  I would guess
that the answer is "you can do this, but that doesn't mean you should."

I'm
  replying
    via
      gmane --
         it
           complains
              if the
                quoted material
                    is too large
                       a fraction of my
                         post.
                           Sorry.
       
   Ben Bolker

______________________________________________
[hidden email] 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.