R help - May 2007 - Function to Sort and test AIC for mixed model lme?

Hi List

I'm running a series of mixed models using lme, and I wonder if there  
is a way to sort them by AIC prior to testing using anova 
(lme1,lme2,lme3,....lme7) other than by hand.

My current output looks like this.

anova 
(lme.T97NULL.ml,lme.T97FULL.ml,lme.T97NOINT.ml,lme.T972way.ml,lme.T97fc. 
ml, lme.T97ns.ml, lme.T97min.ml)


                 Model df      AIC      BIC    logLik   Test    
L.Ratio p-value
lme.T97NULL.ml      1  3 624.3053 635.0671 -309.1527
lme.T97FULL.ml      2 14 508.7594 558.9809 -240.3797 1 vs 2  
137.54588  <.0001
lme.T97NOINT.ml     3  7 564.8574 589.9682 -275.4287 2 vs 3   
70.09798  <.0001
lme.T972way.ml      4 12 507.0991 550.1461 -241.5496 3 vs 4   
67.75832  <.0001
lme.T97fc.ml        5  9 502.5878 534.8730 -242.2939 4 vs 5    
1.48869  0.6849
lme.T97ns.ml        6  8 504.8971 533.5951 -244.4485 5 vs 6    
4.30928  0.0379
lme.T97min.ml       7  6 569.8188 591.3423 -278.9094 6 vs 7   
68.92177  <.0001

Thanks

Ken



	[[alternative HTML version deleted]]

Ken Nussear <knussear <at> mac.com> writes:
> I'm running a series of mixed models using lme, and I wonder if there  
> is a way to sort them by AIC prior to testing using anova 
> (lme1,lme2,lme3,....lme7) other than by hand.
You can try something like the following. However, also consider using 
dropterm or stepAIC in MASS.

Dieter


#---------------------
library(nlme)
fmlist = vector("list",2)
fmlist[[1]] = lme(distance ~ age, data = Orthodont,method="ML") 
fmlist[[2]] = lme(distance ~ age + Sex, data = Orthodont, 
  random = ~ 1,method="ML")
aic0 = unlist(lapply(fmlist,AIC))
aic0 # larger first
fmlist1 = fmlist[order(aic0)]
unlist(lapply(fmlist1,AIC)) # smaller first

> Ken Nussear <knussear <at> mac.com> writes:
>
> > I'm running a series of mixed models using lme, and I wonder if  
> there
> > is a way to sort them by AIC prior to testing using anova
> > (lme1,lme2,lme3,....lme7) other than by hand.
>
> You can try something like the following. However, also consider  
> using dropterm or stepAIC in MASS.
>
> Dieter
>
> #---------------------
>
> library(nlme)
> fmlist = vector("list",2)
> fmlist[[1]] = lme(distance ~ age, data = Orthodont,method="ML")  
> fmlist[[2]] = lme(distance ~ age + Sex, data = Orthodont,   random  
> = ~ 1,method="ML")
> aic0 = unlist(lapply(fmlist,AIC))
> aic0 # larger first
> fmlist1 = fmlist[order(aic0)]
> unlist(lapply(fmlist1,AIC)) # smaller first
I looked at stepAIC, but wanted to specify my own models.

I have come pretty close with this

# grab all lme objects
tst1 <- ls(pat=".ml")
 > tst1
[1] "lme.T972way.ml"  "lme.T97FULL.ml" 
"lme.T97NOINT.ml"
"lme.T97NULL.ml"  "lme.T97fc.ml"   
"lme.T97min.ml"
[7] "lme.T97ns.ml"

#create array of AIC for all in tst1
tst2 <- lapply(lapply(tst1,get),AIC)
 > tst2
[[1]]
[1] 507.0991

[[2]]
[1] 508.7594

[[3]]
[1] 564.8574

[[4]]
[1] 624.3053

[[5]]
[1] 502.5878

[[6]]
[1] 569.8188

[[7]]
[1] 504.8971

#sort tst1 by order of tst2
tst3 <- tst1[order(unlist(tst2))]

 > tst3
[1] "lme.T97fc.ml"    "lme.T97ns.ml"   
"lme.T972way.ml"
"lme.T97FULL.ml"  "lme.T97NOINT.ml"
"lme.T97min.ml"
[7] "lme.T97NULL.ml"


The problem comes with getting the final anova statement to run.

 >anova(tst3)
Error in anova(tst3) : no applicable method for "anova"

I also tried

tst4 <- paste(toString(tst3),collapse="")

 > tst4
[1] "lme.T97fc.ml, lme.T97ns.ml, lme.T972way.ml, lme.T97FULL.ml,  
lme.T97NOINT.ml, lme.T97min.ml, lme.T97NULL.ml"
 >
 > anova(tst4)
Error in anova(tst4) : no applicable method for "anova"

Any ideas on getting the last part to work?






	[[alternative HTML version deleted]]

Hi List,
Is there a package in R to find L1-SVM. I did search and found svmpath
but not sure if it is what I need.
Thanks,
Saeed

> Is there a package in R to find L1-SVM. I did search and found svmpath
> but not sure if it is what I need.

Try kernlab