[R-sig-ME] looking for differencies in G structure among factor levels using MCMCglmm

[Diego Carmona]

 I am working with a multi-trait MCMCglmm model. The model has four plant
traits,
a three levels treatment is considered as a “fixed” factor and, as a random
factor,
I included the term family (64 full maternal siblings).
What I am willing to know is the syntax that allow me to estimate the
genetic variance-covariance matrices
for each treatment.

In general, I am looking for differences in G structure among factor
levels. I guess that it should be
something like

G-structure: ~ us (trait):fam
             ~ us (trait):fam:tratA
          ~ us (trait):fam:tratB
             ~ us (trait):fam:tratC

R-structure: ~us(trait):units



This was my last attempt.

prior2<-list(G=list(G1=list(V=phen.var1/4,n=2),
                      G2=list(V=phen.var1/4,n=2)),
                      R=list(V=phen.var1/4,n=2))

bayes.global1<-MCMCglmm(cbind(crece,
R,flor,frutos)~trait+trat+trait:trat-1,random =
~us(trait):fam+us(trait):fam:trat2,
rcov = ~us(trait):units, family = c("gaussian", "gaussian","gaussian",
"poisson"),
data = na.omit(basep), prior = prior2, verbose = FALSE,singular.ok=TRUE,
nitt=13000, thin=10, burnin=3000)

This is part of the summary.


 G-structure:  ~us(trait):fam

                   post.mean   l-95% CI  u-95% CI eff.samp
crece:crece.fam    0.0043452  0.0015103 8.050e-03   1000.0
R:crece.fam       -0.0002719 -0.0014696 9.586e-04   1000.0
flor:crece.fam     0.0436852 -0.0689005 1.670e-01   1000.0
frutos:crece.fam   0.0053743 -0.0247571 4.286e-02   1000.0
etc...

    ~us(trait):fam:trat2

                         post.mean   l-95% CI  u-95% CI eff.samp
crece:crece.fam:trat2    0.0039826  0.0015898 6.967e-03    655.9
R:crece.fam:trat2       -0.0001361 -0.0010399 8.509e-04    900.4
flor:crece.fam:trat2     0.0237493 -0.0580506 1.238e-01    588.7
frutos:crece.fam:trat2   0.0046849 -0.0135320 2.080e-02   1000.0
etc...

 R-structure:  ~us(trait):units

                     post.mean  l-95% CI  u-95% CI eff.samp
crece:crece.units    0.0454059  0.038990  0.051924    848.7
R:crece.units       -0.0008172 -0.002948  0.001473   1000.0
flor:crece.units    -0.0084173 -0.201620  0.199751   1000.0
frutos:crece.units   0.0729428  0.048952  0.096300   1000.0
crece:R.units       -0.0008172 -0.002948  0.001473   1000.0


Location effects: cbind(crece, R, flor, frutos) ~ trait + trat2 +
trait:trat2 - 1

                   post.mean  l-95% CI  u-95% CI eff.samp  pMCMC
traitcrece          0.708005  0.660129  0.754760     1000 <0.001 ***
traitR              1.310004  1.287974  1.332702     1000 <0.001 ***
traitflor          56.262578 54.290430 58.332811     1000 <0.001 ***
traitfrutos         1.675350  1.189332  2.296322     1000 <0.001 ***
trat22              0.005729 -0.057418  0.058968     1000  0.810
trat23              0.014290 -0.045559  0.070862     1000  0.652
traitR:trat22      -0.134281 -0.199170 -0.061456     1421 <0.001 ***
traitflor:trat22   -3.233360 -5.507265 -0.632871     1263  0.014 *
traitfrutos:trat22  0.008789 -0.417274  0.505621     1000  0.998
traitR:trat23      -0.111668 -0.179654 -0.047945     1000 <0.001 ***
traitflor:trat23   -0.210391 -2.620701  2.472176     1000  0.912
traitfrutos:trat23 -0.145282 -0.724167  0.288831     1000  0.570

Many thanks

Muchas Gracias

	[[alternative HTML version deleted]]