[R-sig-ME] Interpreting a MCMCglmm model with a bivariate response variable
Jarrod Hadfield
j.hadfield at ed.ac.uk
Wed Sep 23 16:25:02 CEST 2015
Hi Iker,
You need to follow the advice given to your previous post. With
unconstrained residual variance the model is largely generating
nonsense. Use `corg' instead of `us'. Also, depending on what the
outcomes are you almost certainly need to have `trait' in the fixed
effect specification.
Jarrod
Quoting Iker Vaquero Alba <karraspito at yahoo.es> on Wed, 23 Sep 2015
13:28:09 +0000 (UTC):
>
> Hello all,
> I am implementing a model with a multiple (bivariate) response
> variable using MCMCglmm. Both response variables and all the
> explanatory variables are categorical variables, with between 2 and
> 6 levels. The model is as follows:
>
> testmodel1<-MCMCglmm(cbind(natapshort,nataplong)~gender+age+religion+sexor+selfattr+partnerattr+gender:age+gender:religion+gender:sexor+gender:selfattr+gender:partnerattr+age:religion+age:sexor+age:selfattr+age:partnerattr+religion:sexor+religion:selfattr+religion:partnerattr+sexor:selfattr+sexor:partnerattr+selfattr:partnerattr,random=NULL,rcov=~us(trait):units,family=c("threshold","threshold"),data=extphen,nitt=100000,singular.ok=TRUE)
> And this is the summary of the model after all the iterations:
>
> summary(testmodel1) Iterations =3001:99991 Thinninginterval =
> 10 Sample size = 9700 DIC: R-structure:
> ~us(trait):units post.mean l-95% CI u-95%
> CI eff.sampnatapshort:natapshort.units 114108 992.1 213000
> 7.105nataplong:natapshort.units 33245 310.8 66300
> 4.656natapshort:nataplong.units 33245 310.8 66300
> 4.656nataplong:nataplong.units 82964 671.5 155869
> 2.218 Location effects:cbind(natapshort, nataplong) ~ gender + age
> + religion + sexor + selfattr +partnerattr + gender:age +
> gender:religion + gender:sexor + gender:selfattr +gender:partnerattr
> + age:religion + age:sexor + age:selfattr + age:partnerattr+
> religion:sexor + religion:selfattr + religion:partnerattr +
> sexor:selfattr +sexor:partnerattr + selfattr:partnerattr
> post.mean l-95% CI u-95% CI eff.samp
> pMCMC (Intercept) 8.934e+02 -6.995e+02 2.596e+03
> 275.73 0.22660 genderM -8.936e+01 -7.437e+0
> 2 5.066e+02 4236.67 0.75794 genderO -1.292e+03
> -1.934e+05 1.765e+05 9700.00 0.99052
> age -3.493e+02 -1.170e+03 3.615e+02 505.92
> 0.31918 religionY -7.361e+02 -1.598e+03 1.481e+01
> 33.71 0.03402 * sexorHOM -1.235e+03
> -1.808e+05 1.679e+05 9700.00 0.99814 sexorOT
> 2.193e+03 -1.589e+05 1.687e+05 10583.09 0.97814
> selfattr -2.367e+02 -7.424e+02 1.706e+02 314.82
> 0.24948 partnerattr 1.391e+02 -2.667e+02 6.147e+02
> 966.40 0.49546 genderM:age 2.696e+01 -1.313e+02
> 1.748e+02 3786.10 0.69670 genderO:age -1.055e+04
> -8.325e+04 7.163e+04 1722.63 0.78474
> genderM:religionY -1.295e+02 -3.725e+02 8.194e+01 200.08
> 0.20495 genderO:religionY -1.016e+04 -1.589e+05 1.505e+05
> 8731.86 0.89052 genderM:sexorHOM -2.245e+02
> -5.713e+02 4.443e+01 105.67 0.10495
> genderO:sexorHOM -8.104e+03 -1.620e+05 1.385e+05 5318.22
> 0.90474 genderM:sexorOT -1.52
> 0e+02 -5.124e+02 1.856e+02 423.76 0.33402
> genderO:sexorOT 2.628e+03 -1.654e+05 1.658e+05 9700.00
> 0.97670 genderM:selfattr 9.029e+01 -3.152e+01 2.334e+02
> 119.78 0.12907 genderO:selfattr 6.281e+03
> -6.511e+04 8.524e+04 3504.67 0.88412
> genderM:partnerattr -7.284e+01 -2.160e+02 6.729e+01 263.29
> 0.25052 genderO:partnerattr 2.536e+02 -5.113e+02 1.121e+03
> 1291.76 0.49113 age:religionY 8.732e+01
> -1.283e+02 3.457e+02 727.46 0.42289
> age:sexorHOM 2.809e+02 -8.592e+04 8.847e+04 9700.00
> 0.99711 age:sexorOT -1.246e+03 -8.447e+04 7.941e+04
> 9370.57 0.97526 age:selfattr 1.195e+02
> -6.636e+01 3.452e+02 212.35 0.19567
> age:partnerattr -8.598e+00 -1.963e+02 1.714e+02 9700.00
> 0.92227 religionY:sexorHOM 8.506e+01 -2.392e+02 4.612e+02
> 2059.92 0.59959 religionY:sexorOT 1.420e+01
> -5.170e+02 5.464e+02 9700.00 0.96268
> religionY:selfattr 2.782e+01 -1.198e+02 1.833e+02 3520.80
> 0.68701 religionY:part
> nerattr 1.407e+02 1.423e+01 2.886e+02 22.99 0.00928
> **sexorHOM:selfattr 1.160e+02 -1.141e+02 3.707e+02 394.74
> 0.28495 sexorOT:selfattr 1.006e+02 -8.528e+01 3.050e+02
> 305.50 0.24577 sexorHOM:partnerattr 1.231e+02
> -1.246e+02 3.990e+02 415.43 0.31072
> sexorOT:partnerattr -1.401e+00 -2.007e+02 1.956e+02 9700.00
> 0.99237 selfattr:partnerattr 5.483e+00 -6.017e+01 7.207e+01
> 3007.45 0.85464 ---Signif. codes: 0?***? 0.001 ?**? 0.01 ?*? 0.05
> ?.? 0.1 ? ? 1 Cutpoints: post.mean l-95%
> CI u-95% CI eff.sampcutpoint.traitnatapshort.1 235.2 62.34
> 376.2 8.822cutpoint.traitnatapshort.2 633.4 202.16
> 944.0 3.578cutpoint.traitnatapshort.3 1139.5 364.35
> 1683.7 5.832cutpoint.traitnataplong.1 293.4 54.85
> 433.7 5.203cutpoint.traitnataplong.2 651.8 223.70
> 961.6 2.604cutpoint.traitnataplong.3 1023.1 344.82
> 1483.0 2.353
>
> So, my question is: in that summary, where are the effect sizes, are
> they the "post. mean" column? And have they been transformed in some
> way? Because obviously, for response variables that can only take
> values 1,2,3,4 or 5, I would expect to see those as the effect size.
> Also, is there any way of knowing to what extent are those results
> due to each specific response variable, and the degree of covariance
> between both? Is it possible to get all that information from that
> summary output I have copied above?
> Thank you very much. Iker
>
> __________________________________________________________________
>
> Iker Vaquero-Alba
> Visiting Postdoctoral Research Associate
> Laboratory of Evolutionary Ecology of Adaptations
> Joseph Banks Laboratories
> School of Life Sciences
> University of Lincoln Brayford Campus, Lincoln
> LN6 7DL
> United Kingdom
>
> https://eric.exeter.ac.uk/repository/handle/10036/3381
>
>
> [[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.
More information about the R-sig-mixed-models
mailing list