R help - Jun 2012 - MuMIn - assessing variable importance following model averaging, z-stats/p-values or CI?

Dear R users,

Recent changes to the MuMIn package now means that the model averaging command
(model.avg) no longer returns confidence intervals, but instead returns zvalues
and corresponding pvalues for fixed effects included in models.

Previously I have used this package for model selection/averaging following
Greuber et al (2011) where it suggests that one should use confidence intervals
from model averaging to assess whether your fixed effects have an affect or not 
(If confidence intervals do not span zero then variable has an affect).

Can anyone tell me why MuMIn now gives z-stats and p-values and whether these
should be used to assess the 'significance'/importance of variables when
model averaging?

Heres the example code of what I'm doing....

#-------------------------------------------------------------------------------------#
ps<-lmer(tranPS~(
    Sex+
    Age.Cat2+
    TOTAL+
    Propfarm+
    Maize+
    TOTAL:Propfarm+
    Maize:TOTAL+
    Maize:Propfarm+
    (1|Socialgroup)+(1|Year)+(1|Tattoo)),REML=FALSE, data=propspec)

pss<-standardize(ps,standardize.y = FALSE)

psdrg<-dredge(pss)

summary(model.avg(get.models(psdrg,subset=delta<2)))
#-------------------------------------------------------------------------------------#

REf -Grueber, C.E., Nakagawa, S., Laws, R.J. & Jamieson, I.G. (2011)
Multimodel inference in ecology and evolution: challenges and solutions. Journal
of evolutionary biology, 24, 699-711.

Any help would be much appreciated

Regards

Andrew Robertson
PhD student
Centre for Ecology and Conservation
University of Exeter, Cornwall Campus
Tremough, Cornwall. TR10 9EZ
UK
Tel: 01326 371852
Email: ar313 at exeter.ac.uk
Web page:
http://biosciences.exeter.ac.uk/staff/postgradresearch/andrewrobertson/

Please contact the package maintainer.

Best,
Uwe Ligges

On 26.06.2012 12:46, Robertson, Andrew wrote:
> Dear R users,
>
> Recent changes to the MuMIn package now means that the model averaging
command (model.avg) no longer returns confidence intervals, but instead returns
zvalues and corresponding pvalues for fixed effects included in models.
>
> Previously I have used this package for model selection/averaging following
Greuber et al (2011) where it suggests that one should use confidence intervals
from model averaging to assess whether your fixed effects have an affect or not 
(If confidence intervals do not span zero then variable has an affect).
>
> Can anyone tell me why MuMIn now gives z-stats and p-values and whether
these should be used to assess the 'significance'/importance of
variables when model averaging?
>
> Heres the example code of what I'm doing....
>
>
#-------------------------------------------------------------------------------------#
> ps<-lmer(tranPS~(
>      Sex+
>      Age.Cat2+
>      TOTAL+
>      Propfarm+
>      Maize+
>      TOTAL:Propfarm+
>      Maize:TOTAL+
>      Maize:Propfarm+
>      (1|Socialgroup)+(1|Year)+(1|Tattoo)),REML=FALSE, data=propspec)
>
> pss<-standardize(ps,standardize.y = FALSE)
>
> psdrg<-dredge(pss)
>
> summary(model.avg(get.models(psdrg,subset=delta<2)))
>
#-------------------------------------------------------------------------------------#
>
> REf -Grueber, C.E., Nakagawa, S., Laws, R.J. & Jamieson, I.G. (2011)
Multimodel inference in ecology and evolution: challenges and solutions. Journal
of evolutionary biology, 24, 699-711.
>
> Any help would be much appreciated
>
> Regards
>
> Andrew Robertson
> PhD student
> Centre for Ecology and Conservation
> University of Exeter, Cornwall Campus
> Tremough, Cornwall. TR10 9EZ
> UK
> Tel: 01326 371852
> Email: ar313 at exeter.ac.uk
> Web page:
http://biosciences.exeter.ac.uk/staff/postgradresearch/andrewrobertson/
> ______________________________________________
> 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.
>

p-value > 0.05 means that the 95% confidence intervals span zero.
Use "confint" to get the CI. It is described in
"?model.avg".

cheers,
kamil


Dnia 2012-06-27 12:00, Robertson, Andrew pisze:
> Dear R users,
>
> Recent changes to the MuMIn package now means that the model averaging
command (model.avg) no longer returns confidence intervals, but instead returns
zvalues and corresponding pvalues for fixed effects included in models.
>
> Previously I have used this package for model selection/averaging following
Greuber et al (2011) where it suggests that one should use confidence intervals
from model averaging to assess whether your fixed effects have an affect or not 
(If confidence intervals do not span zero then variable has an affect).
>
> Can anyone tell me why MuMIn now gives z-stats and p-values and whether
these should be used to assess the 'significance'/importance of
variables when model averaging?
<...>