Skip to content
Prev 1065 / 7420 Next

nested mixed model?

Hi Mauro,

I agree with both Thierry and Luciano, considering your nested design the random effects lie at the level of the individual.  My only concern is that given the coding below, only the intercept varies and the slopes are set to be equal for the random variables.  From what I have read (e.g. Schielzeth and Forstmeier 2008, Behavioral Ecology), not investigating the slopes of the random variables can lead to spurious results.  I would also consider using Markov Chain Monte Carlo (MCMC) approximations (see Bolker et al. 2009, TREE).  However, I don't know if MCMC estimations work with nested designs.  Hence, perhaps the following might work?


tooth.lmer<-lmer(response ~ species, data=tooth, random=~age_of_individual|individual/bone/tooth)
tooth.pval<-pvals.fnc(tooth.lmer, nsim=1000, withMCMC=TRUE)
tooth.pval$fixed
tooth.pval$random


Perhaps others have an opinion regarding variation of slopes and intercepts with mixed effects models?  And the use of MCMC estimations with nested designs?

All the best,

Andrew



Andrew Kosydar
University of Washington
Department of Biology
24 Kincaid Hall, Box 351800
Seattle, WA 98195
USA
On Wed, 3 Feb 2010, ONKELINX, Thierry wrote: