[R-sig-ME] distribution for ratios?

[i white]

Treat as Gaussian after logistic or arcsine-square root transformation?

On 01/23/2014 10:07 AM, Celine Teplitsky wrote:
> Dear all,
>
> we are trying to run a meta-analysis on factors affecting heritability
> in MCMCglmm. As it is a ratio of 2 variances, the values are bounded
> between one and zero, but can not be interpreted as Success/NbTrials, so
> we can not use a binomial distribution.
>
> So the trouble we're having is knowing which distribution to use? Should
> we just use a Gaussian?
>
> Many thanks in advance for your help
>
> Best regards
>
> Celine
>

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.