[R] R's AIC values differ from published values

Bert Gunter gunter.berton at gene.com
Mon Feb 13 20:26:44 CET 2012


This is answered in ?AIC. Have you read it?

-- Bert

On Mon, Feb 13, 2012 at 10:22 AM, david hamer <j.david.hamer at gmail.com> wrote:
> Using the Cement hardening data in Anderson (2008) Model Based Inference in
> the Life Sciences. A Primer on Evidence, and working with the best model
> which is
>    lm  (  y  ~   x1   +   x2,    data = cement  )
> the AIC value from R is
>
>    model    <-    lm  ( formula   =   y   ~   x1   +  x2  ,   data =
> cement  )
>    AIC ( model )
>
> 64.312
>
> which can be converted to AICc by adding the bias correction factor
> 2*K*(K+1)/(n-K-1) to give the AICc value of
> 69.312
> (addition of 5, where n=13 and K=4).
>
> This same value, 69.31, can be obtained using R package
> AICcmodavg
>
>   library  (  AICcmodavg  )
>   data (cement)
>   cement
>   Cand.models   <-   list( )
>   Cand.models[[1]]   <-   lm  (  y  ~   x1   +   x2,    data = cement  )
>   Cand.models[[2]]   <-   lm  (  y    ~   x3   +   x4,     data = cement  )
>   Cand.models[[3]]   <-   lm  (  y   ~   x1 + x2 + x1 * x2,   data =
> cement  )
>   Cand.models[[4]]   <-   lm  (  y   ~   x3  +  x4  +  x3 * x4,   data =
> cement  )
>     ##   vector of model names
>   Modnames   <-   paste("MODEL", 1:4, sep="     ")
>     ##   AICc
>   aictab   (  cand.set   =   Cand.models,   modnames  =  Modnames  )
>
> However, the AICc value reported by Anderson (2008) is
> 32.41.
> The AICc value obtained using RSS value (i.e., calculating AICc "manually"
> from the output of linear regression) is
> 32.41.
>
> Thanks for any help.
> David
> New R user, minimal familiarity with statistics.
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.



-- 

Bert Gunter
Genentech Nonclinical Biostatistics

Internal Contact Info:
Phone: 467-7374
Website:
http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm



More information about the R-help mailing list