AIC for model selection in Package ‘MuMIn’

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

AIC for model selection in Package ‘MuMIn’

elaine kuo
Dear List,

Ben's recommendation to MuMIn is of great importance to research with over
ten independent variables.

However, the criteria in this package for model selection are AICc
(second-order AIC) and QAIC (quasi-AIC).
Please kindly share if it is possible to carry AIC rather than the metrics
above using this package.
Thank you.

Elaine

        [[alternative HTML version deleted]]

______________________________________________
[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: AIC for model selection in Package �MuMIn�

bbolker
elaine kuo <elaine.kuo.tw <at> gmail.com> writes:

[re: MuMIn package]

> However, the criteria in this package for model selection are AICc
> (second-order AIC) and QAIC (quasi-AIC).
> Please kindly share if it is possible to carry AIC rather than the metrics
> above using this package.
> Thank you.

  I think you're misunderstanding the documentation.  The reason
that AICc and QAIC are explicitly included in the package, and the
package documentation, is that (unlike AIC and BIC) they are *not*
implemented in base R.  If you simply use dredge(...,rank="AIC")
that should work.

  good luck.

______________________________________________
[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: AIC for model selection in Package �MuMIn�

elaine kuo
Thanks for Ben.

Model selection by AIC is feasible by the code below.
One more question, when is it appropriate to carry out model averaging and
parameter averaging?

code
library(MuMIn)
data(Cement)
lm1  <-  lm(y  ~  .,  data  =  Cement)
dd  <-  dredge(lm1,  beta  = TRUE,  eval  =  TRUE,  rank  =  "AIC")
print(dd)

Elaine

On Mon, Aug 16, 2010 at 9:21 PM, Ben Bolker <[hidden email]> wrote:

> elaine kuo <elaine.kuo.tw <at> gmail.com> writes:
>
> [re: MuMIn package]
>
> > However, the criteria in this package for model selection are AICc
> > (second-order AIC) and QAIC (quasi-AIC).
> > Please kindly share if it is possible to carry AIC rather than the
> metrics
> > above using this package.
> > Thank you.
>
>  I think you're misunderstanding the documentation.  The reason
> that AICc and QAIC are explicitly included in the package, and the
> package documentation, is that (unlike AIC and BIC) they are *not*
> implemented in base R.  If you simply use dredge(...,rank="AIC")
> that should work.
>
>  good luck.
>
> ______________________________________________
> [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.
>

        [[alternative HTML version deleted]]

______________________________________________
[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: AIC for model selection in Package â??MuMInâ??

bbolker
   That's really too big a question to do justice to in a mailing list
reply. If possible, consult a local expert.

   To keep it very short, I would say:
  (1) when you are interested in maximizing expected predictive accuracy
(rather than testing hypotheses);
  (2) when the parameters in the models are commensurate (e.g., be
careful comparing models with and without interaction terms). When in doubt,
figure out whether the averaged parameter can be interpreted as a
concrete prediction about some experimental or observational contrast.

elaine kuo wrote:

> Thanks for Ben.
>
> Model selection by AIC is feasible by the code below.
> One more question, when is it appropriate to carry out model averaging
> and parameter averaging?
>
> code
> library(MuMIn)
> data(Cement)
> lm1  <-  lm(y  ~  .,  data  =  Cement)
> dd  <-  dredge(lm1,  beta  = TRUE,  eval  =  TRUE,  rank  =  "AIC")
> print(dd)
>
> Elaine
>
> On Mon, Aug 16, 2010 at 9:21 PM, Ben Bolker <[hidden email]
> <mailto:[hidden email]>> wrote:
>
>     elaine kuo <elaine.kuo.tw <http://elaine.kuo.tw> <at> gmail.com
>     <http://gmail.com>> writes:
>
>     [re: MuMIn package]
>
>     > However, the criteria in this package for model selection are AICc
>     > (second-order AIC) and QAIC (quasi-AIC).
>     > Please kindly share if it is possible to carry AIC rather than
>     the metrics
>     > above using this package.
>     > Thank you.
>
>      I think you're misunderstanding the documentation.  The reason
>     that AICc and QAIC are explicitly included in the package, and the
>     package documentation, is that (unlike AIC and BIC) they are *not*
>     implemented in base R.  If you simply use dredge(...,rank="AIC")
>     that should work.
>
>      good luck.
>
>     ______________________________________________
>     [hidden email] <mailto:[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.
>
>

______________________________________________
[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.