[R-sig-ME] mcmcsamp in lme4

Ben Bolker bolker at ufl.edu
Wed Jun 10 00:18:33 CEST 2009


   There is indeed a simulate.lme() function in nlme:
can you use this to construct a Monte Carlo simulation?
i.e. (pseudocode)

 fit full model (model1)
 fit appropriate reduced model (model0)

 N times (using replicate() or a for loop) {
     simulate from reduced model, using simulate.lme(model0);
         pack it back into a copy of the original data frame
     fit full model [in lme4 there is a "refit" method that is
        more efficient, but apparently not in nlme]
            to simulated (null data)
     fit reduced model (minus one or more fixed terms), ditto
     compute/save deviance difference between full and reduced model
       as results[i]
}

compare observed deviance difference to simulated distribution:

 mean(c(results,obsdev)>=obsdev)

  cheers
    Ben


De Smedt Sebastiaan wrote:
> I fitted a mixed-effects model with a Variance Structure in nlme
> (because Variance Structures are not yet implemented in lme4). After
> several model fittings (with method="ML"), I found the optimal fixed
> structure of my explanatory variables. 
> 
> I understood from this mailing list that the p-values of the fixed
> effects, given by summary(namemodel) are not trustable and thus I want
> to use a Monte Carlo simulation in order to estimate the significance of
> the fixed terms. This command doesn't exist in nlme, and I can't fit my
> model in lme4. Is there a way to overcome this problem? Or another
> function in nlme that gives comparable results?
> 
>  
> 
>  
> 
> Thanks a lot,
> 
> Sebastiaan
> 
>  
> 
> Sebastiaan De Smedt
> 
> Research assistant
> 
> UA - Department of Bioscience Engineering
> 
>  
> 
> Groenenborgerlaan 171 - V6.16
> 
> B-2020 Antwerpen, Belgium
> 
> +32 (0)3 265 34 51
> 
>  
> 
>  
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models


-- 
Ben Bolker
Associate professor, Biology Dep't, Univ. of Florida
bolker at ufl.edu / www.zoology.ufl.edu/bolker
GPG key: www.zoology.ufl.edu/bolker/benbolker-publickey.asc




More information about the R-sig-mixed-models mailing list