[R-sig-ME] Wald test on heritabilty estimates from MCMCglmm

[Jarrod Hadfield]

Dear Ardo,

I think the Wald test will lead you to draw incorrect conclusions for  
bounded posteriors that are highly skewed. I think a priori that if  
the trait is variable then the probability that h2 = 0 is small  
(considerably smaller than the p-value everyone is searching for). For  
me at least, the important thing is to say something accurate and  
precise about the magnitude of h2, and this information is neatly  
summarised using HPDinterval. It is common to see h2 estimates that  
are "significant" but the CI's are essentially 0.01 to 1.   Such  
studies tell us little that we did not already know: a proportion lies  
between 0 and 1.

Cheers,

Jarrod


On 12 Oct 2010, at 15:22, Ndjido Ardo BAR wrote:

> Hi Folks!
> I would like to performe a Wald test on heritabilities I estimated  
> from
> MCMCglmm. Does it make sense to use the variance posterieur of h² to  
> build
> Wald statistics (h²/sqrt(Var_post))?
> Ardo
>
> 	[[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.