Message-ID: <13524669.ijSL9NVrgx@flyosfixe>
Date: 2018-11-07T18:09:42Z
From: Pierre de Villemereuil
Subject: MCMCglmm: square root of the sampling variance of additive genetic variance
In-Reply-To: <CAJtCY7U9SsBjNrwbJ=5ytYj_RLU+fdjZ5ZO6ZyxJUNE9qsCj4Q@mail.gmail.com>
Hi,
Just to clarify this:
> The MCMCglmm estimates the posterior distribution of the additive and
> residual variances, to my knowledge then, there is no standard error
> associated with it rather you can output the HPD interval or highest
> posterior density interval at a 95 or 98% confidence interval, which you
> already have done. I stand to be corrected though.
The sd() of the MCMC chain is indeed an estimate of the standard error (+ some error itself coming from the Monte Carlo process), though HPD interval is indeed probably a better (and more Bayesian, in a way) way to measure the uncertainty.
I might also stand to be corrected! ;)
Cheers,
Pierre.