R help - Oct 2006 - lmer BIC changes between output and anova

list,

i am using lmer to fit multilevel models and trying to use anova to compare the
models.  however, whenever i run the anova, the AIC, BIC and loglik are
different from the original model output- as below.  can someone help me out
with why this is happening?  (i'm hoping the output assocaited with the
anova is right!).

thank you,

darren
> unconditional<-lmer(log50 ~ 1 + (1 | Stream:Site) + (1|Stream), data)
> summary(unconditional)
Linear mixed-effects model fit by REML 
Formula: log50 ~ 1 + (1 | Stream:Site) + (1 | Stream) 
   Data: data 
    AIC    BIC logLik MLdeviance REMLdeviance
 -138.8 -132.8  72.42     -150.4       -144.8

> nosection<-lmer(log50 ~ 1 + meanlogATS + residuallogATS + (1 |
Stream:Site) + (1|Stream), data)
> summary(nosection)
Linear mixed-effects model fit by REML 
Formula: log50 ~ 1 + meanlogATS + residuallogATS + (1 | Stream:Site) +      (1 |
Stream)
   Data: data 
    AIC    BIC logLik MLdeviance REMLdeviance
 -140.8 -130.7   75.4     -168.9       -150.8

> anova(unconditional, nosection)
Data: data
Models:
unconditional: log50 ~ 1 + (1 | Stream:Site) + (1 | Stream)
nosection: log50 ~ 1 + meanlogATS + residuallogATS + (1 | Stream:Site) + 
unconditional:     (1 | Stream)
              Df      AIC      BIC   logLik  Chisq Chi Df Pr(>Chisq)    
unconditional  3 -144.370 -138.294   75.185                             
nosection      5 -158.861 -148.734   84.430 18.491      2  9.657e-05 ***

The two models that you compare differ in their fixed effects. As far as lme had
been concerned, the anova gave a warning because one should not compare two
models differing in their fixed effects if they had been estimated with REML
(Pinheiro & Bates, 2000).

I guess the issue is still the same using lmer (but anyone please correct me, if
I am wrong) but perhaps the warning message is currently not implemented.

Regards, Lorenz
- 
Lorenz Gygax
Centre for proper housing of ruminants and pigs
Swiss Federal Veterinary Office
Agroscope Reckenholz-T?nikon Research Station ART

> -----Original Message-----
> From: r-help-bounces at stat.math.ethz.ch
> [mailto:r-help-bounces at stat.math.ethz.ch]On Behalf Of Darren M. Ward
> Sent: Thursday, October 05, 2006 5:18 PM
> To: r-help at stat.math.ethz.ch
> Subject: [R] lmer BIC changes between output and anova
> 
> list,
> 
> i am using lmer to fit multilevel models and trying to use 
> anova to compare the models.  however, whenever i run the 
> anova, the AIC, BIC and loglik are different from the 
> original model output- as below.  can someone help me out 
> with why this is happening?  (i'm hoping the output 
> assocaited with the anova is right!).
> 
> thank you,
> 
> darren
> 
> > unconditional<-lmer(log50 ~ 1 + (1 | Stream:Site) + 
> (1|Stream), data)
> > summary(unconditional)
> Linear mixed-effects model fit by REML 
> Formula: log50 ~ 1 + (1 | Stream:Site) + (1 | Stream) 
>    Data: data 
>     AIC    BIC logLik MLdeviance REMLdeviance
>  -138.8 -132.8  72.42     -150.4       -144.8
> 
> 
> > nosection<-lmer(log50 ~ 1 + meanlogATS + residuallogATS + 
> (1 | Stream:Site) + (1|Stream), data)
> > summary(nosection)
> Linear mixed-effects model fit by REML 
> Formula: log50 ~ 1 + meanlogATS + residuallogATS + (1 | 
> Stream:Site) +      (1 | Stream) 
>    Data: data 
>     AIC    BIC logLik MLdeviance REMLdeviance
>  -140.8 -130.7   75.4     -168.9       -150.8
> 
> 
> > anova(unconditional, nosection)
> Data: data
> Models:
> unconditional: log50 ~ 1 + (1 | Stream:Site) + (1 | Stream)
> nosection: log50 ~ 1 + meanlogATS + residuallogATS + (1 | 
> Stream:Site) + 
> unconditional:     (1 | Stream)
>               Df      AIC      BIC   logLik  Chisq Chi Df 
> Pr(>Chisq)    
> unconditional  3 -144.370 -138.294   75.185                   
>           
> nosection      5 -158.861 -148.734   84.430 18.491      2  
> 9.657e-05 ***
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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.
>