[R] find lowest AIC of a LM

[Ben Bolker]

Fabio Berzaghi <fabe <at> dmu.dk> writes:

> 
> hello all,
> 
> I have a simple linear model with 4/5 variables that I am trying to fit. 
> I would like to find the lowest AIC value with any combination of all 
> the variables. I would like to implement this with a while/for loop. 
> Possibly I would like to generalize this so then I can use it when I 
> have many more variables. I do not want to use step AIC. At the moment I 
> am doing it manually but I would like to automate the process because I 
> have many species. I want to compute all possible combinations with all 
> the variables and know which combination gives the lowest AIC.
> My function looks like this 
> 

  [snip]

  You should really use a data argument, i.e.

 lm(tl~poly(bs0011yme,2)+poly(bt0011yme,2)+poly(ss0011sme,2)+
    poly(st0011sme,2),data=subxy)

(It won't change the answers but it will make everything work
better if you want to post-process the fit, e.g updating or
predicting)

> I cannot seem to find a library that has a function that does this 
> automatically.

  Check out the MuMIn, AICcmodavg, and glmulti packages.