[R-sig-ME] can mixed effects models handle two dependent variables

[Jarrod Hadfield]

Hi  Sarah,

If you're prepared to MCMC it you can use MCMCglmm. Something like:

fit1.cn<-MCMCglmm(cbind(X13c,X15n)~trait+trait:soil+trait:manure-1,  
random=us(trait):site, rcov=~us(trait):units, data=data,  
family=c("gaussian", "gaussian")

where trait indexes columns of the response (X13c & X13n) and units  
indexes rows.

trait+trait:soil+trait:manure-1 fits trait specific intercepts and separate
effects of soil and manure.

us(trait):site fits different site variances for X13c & X13n, together  
with the covariance.

us(trait):units fits different residual variances for X13c & X13n,  
together with the covariance.

You can not fit the heteroscedastic model exactly, but you could fit  
something like

us(trait:soil):units or idh(trait:soil):units in the random effects in  
addition to the residual component - a bit like a nugget effect I guess.

Cheers,

Jarrod


Quoting Jude Phillips <birdlists at gmail.com>:

> Hi,
>
> I am carrying out a stable isotope study, where I have measured stable
> isotopes of nitrogen and carbon in plants at different sites.  I have been
> able to create mixed models for each isotope separately - eg
>
> fit1.c<-lme(X13c~soil+manure, random=~1|site, weights=varIdent(form = ~ 1 |
> soil), data)
> fit1.n<-lme(X15n~soil+manure, random=~1|site, weights=varIdent(form = ~ 1 |
> soil), data)
>
> What I'm wondering is if I can test both stable isotopes at the same time,
> in a MANOVA style analysis.  I can't seem to find any info on this - if
> someone could point me in the direction of some info, or tell me if this
> isn't possible , I'd much appreciate it.
>
> Thanks, Sarah Adams
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>



-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.