[R-sig-ME] MCMCglmm Heterogenous Residual Variance in a Categorial Model

Jarrod Hadfield j.hadfield at ed.ac.uk
Fri Nov 7 21:38:09 CET 2014

Hi Maria,

You are right - for a categorical variable the units variance cannot  
be estimated for either sex. You can safely use rcov=~units and set  
the variance to one in the prior.



Quoting FERNANDEZ Ana maria <s1049258 at sms.ed.ac.uk> on Fri, 7 Nov 2014  
11:18:04 +0000:

> Dear List,
> I was wondering what the recommendation is for structuring residual  
> variances in the context of a categorial model when you have a sex  
> interaction in the random term of the model (where animal is the  
> pedigree structure).
> MCMCglmm course notes have a brief recommendation about sex  
> interaction and allowing the sexes to have different residual  
> variances: "3.4 Heterogenous Residual Variance To be started... In  
> short - if you’ve fitted a sex by dam interaction I would always  
> allow the sexes to have different residual variances. Use  
> rcov=∼idh(sex):units.”
> But the conflict that I am coming across is another recommendation  
> about how the residual term should be- or needs to be- fixed in a  
> categorical model.
> I have a case-control binary response variable and I want to look at  
> the sex effect interaction.
> Let me know if you need more detail.
> Thanks!
> Maria
> Ana Maria Fernández Pujals
> A.Fernandez-2 at sms.ed.ac.uk<mailto:A.Fernandez-2 at sms.ed.ac.uk>

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