[R-sig-ME] Animal model residual value

timothee.bonnet at ieu.uzh.ch timothee.bonnet at ieu.uzh.ch
Tue May 9 12:52:43 CEST 2017

Hi all, 

As a co-author of the van Benthem & al. (2016) paper, let me try and clarify what was meant here.

I agree with Pierre that this is a semantic problem, more than a modelling assumption. 
We defined plasticity as the non-heritable component of a trait (i.e., 1 - heritability ; or the ability of a genotype to produce different phenotypes). This definition is probably animal modeler slang, but I think it conveys the idea that there is no "truly random" phenotypic variation, only numerous environmental influences, observed or unobserved.

The residual variance from an animal model like the one used in our paper would capture the plasticity in response to all environmental influences (including what you could call developmental noise or random phenotypic variability).
If instead you are interested in the plastic response to a specific environmental variable (a "reaction-norm" definition of plasticity), you will probably have to include this variable explicitly in your model (but difficult to tell how without more details). 



Timothée Bonnet
Post-Doctoral researcher

Department of Evolutionary Biology and Environmental Studies
University of Zürich-Irchel
Winterthurerstrasse 190
CH-8057 Zürich

Office Y13-J-34
phone: +41 (0)44 635 47 66
e.mail: timothee.bonnet at ieu.uzh.ch



It depends on what you call "plasticity". Most often, plasticity is defined as the part of phenotypic variability that varies according to environment. Without an experimental settings or environmental replications, it's very hard to distinguish from random phenotypic variability.

I've heard people considering that the environmental variance is a measure of plasticity, but it seems to me that this is a huge assumption that random variability is negligible, especially if you have only 1 environment.


On Monday, 8 May 2017 11:25:59 NZST Walid wrote:
> Hi everyone,
> I have a question on a certain assumption made regarding the 'animal' 
> model when implemented in a quantitative genetic study for a trait. 
> While reading van Benthem et al. (2016), the author mentions that the 
> residual (environmental) value, in the additive partitioning assumed by 
> the model, captures plasticity. Does this assumption always hold? or 
> only in the case where we model the maternal, permanent environment and 
> common environment?
> My question is for the purpose of estimating the plasticity of a fixed 
> heritable life-history trait (occurs only once during individual 
> lifetime). Since there are no explicit methods to estimate individual 
> plasticity in a non-labile trait, I am attempting to see if I can 
> circumvent this by using the 'animal' model based on the assumption 
> mentioned above.
> Thank you

	[[alternative HTML version deleted]]

More information about the R-sig-mixed-models mailing list