Dear R Experts,
Does anyone have advice for how to nest random effects in the MCMCglmm package?
Right now, I am running my models with Individual as a random effect; however,
the 25 individuals in my study are birds taken from 13 different nests, so I
feel that a more accurate model would nest the variable Individual (called RNR
in my code) within the variable Nest ID.
I am trying to run both univariate and bivariate mixed models using the MCMCglmm
package. (My interest is in calculating repeatability estimates using the
univariate models, and in partitioning out the within-individual and
between-individual correlations using the bivariate models.) I am also trying to
include a fixed effect (SMI) but am not sure that I am doing this correctly for
either case.
Example Univariate MM:
Cort0.mcmc<-MCMCglmm(CORT0_LOG10_Z~SMI_CEN, random=~indiv,
family="gaussian", data=cort0.data,verbose=FALSE)
CORT0_LOG10_Z is the variable of interest (which has been measured repeatedly
for each given individual)
RNR is individual ID
SMI_CEN is a fixed effect that varies both within and between individuals
Example piece of data frame:
RNR HPA CORT0_LOG10_Z SMI_CEN NEST_ID
1 AR04273 1 -3.29844317 1.236170087 3
2 AR04273 2 -0.51328002 0.122035902 3
3 AR04281 1 0.83631319 1.212914325 8
Example Bivariate MM:
cort0_cort15.bivarC.mcmc<-MCMCglmm(cbind(CORT0_std,CORT15_std)~(SMI_cen),
random=~us(trait):RNR, rcov=~us(trait):units,
family=c("gaussian","gaussian"), prior=prior.bivar,
nitt=1300000,thin=1000,burnin=300000, data=cort0_cort15,verbose=FALSE)
CORT0_std and CORT15_std are the 2 variables of interest
RNR is individual ID
SMI_cen is a fixed effect that varies both within and between individuals
Example piece of data frame:
RNR SMI SMI_avg SMI_dev CORT0 CORT15 SMI_cen SMI_avg_cen
CORT0_std CORT15_std NEST_ID
1 AR04273 18.03640 17.48987 0.546522102 0.460 4.590 1.236170087 0.73332960
-3.80922656 -3.170123623 3
2 AR04273 16.94335 17.48987 -0.546522102 3.493 13.604 0.122035902 0.73332960
-0.04862913 -0.480133012 3
3 AR04281 15.63367 15.86610 -0.232427804 4.875 13.507 -1.212914325 -1.05393597
0.56974795 -0.497849822 8
I am basing my code off of the code supplied in the 2013 paper by Dingemanse and
Dochtermann about applying linear mixed modelling to behavioral ecology.
Does anyone have advice for how to nest RNR within Nest ID, as well as insight
into whether I have coded the fixed effect SMI correctly?
Many thanks,
Marian