How to use fixed variance components in Multitrait function?

[HONG-genetics]

I would like to do cross validation with fixed variance. How to use fixed variance components in Multitrait function?

[gdlc]

You can probably do that by choosing a prior with very large DF. We use an Inverse Wishart prior, the hyper-parameters are the prior scale (S) and the prior DF. The prior mode is S/(df+p+1) where p is the number of traits. So, if your target variance is V, you can choose. S=V*(1e8+p+1) and df=1e8 that should give you a posterior mean almost identical to V. I am copying Paulino who may remember better the way we parameterize the inverse wishart in Multitrait. Gustavo

…

On Wed, Jun 16, 2021 at 6:35 PM HONG-genetics ***@***.***> wrote: I would like to do cross validation with fixed variance. How to use fixed variance components in Multitrait function? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#55>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABGQ3IUR4BACK2BSV2MRM33TTERJ5ANCNFSM462MPZ5A> .

[gdlc]

The parameterization used for the inverse wishart distribution is the standard one, which can be found in the wiki, Regards.

…

On Fri, Jun 18, 2021 at 7:18 AM gdeloscampos ***@***.***> wrote: You can probably do that by choosing a prior with very large DF. We use an Inverse Wishart prior, the hyper-parameters are the prior scale (S) and the prior DF. The prior mode is S/(df+p+1) where p is the number of traits. So, if your target variance is V, you can choose. S=V*(1e8+p+1) and df=1e8 that should give you a posterior mean almost identical to V. I am copying Paulino who may remember better the way we parameterize the inverse wishart in Multitrait. Gustavo On Wed, Jun 16, 2021 at 6:35 PM HONG-genetics ***@***.***> wrote: > I would like to do cross validation with fixed variance. How to use fixed > variance components in Multitrait function? > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#55>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABGQ3IUR4BACK2BSV2MRM33TTERJ5ANCNFSM462MPZ5A> > . >