lchol_prior: Scale-invariant variance matrix prior on the log-Cholesky...

In mlysy/losmix: Inference for Gaussian Location-Scale Mixed-Effects Models

Description

Scale-invariant variance matrix prior on the log-Cholesky scale.

Usage

lchol_prior(Omega)

Arguments

Omega

Precision matrix parameter. A matrix of size p x p, or an array of size p x p x n.

Details

For a given variance (or precision) matrix Omega, the scale-invariant prior is

pi(Omega) \propto 1/|Omega|

in the sense that the change of variables Psi = A Omega t(A) results in the prior pi(Psi) = 1/|Psi|.

On the log-Cholesky scale defined by log_chol, the scale-invariant prior is given by

pi(logC) \propto prod( diag(chol(Omega))^(p-(1:p)) )

Value

A log-prior vector of length n.

mlysy/losmix documentation built on Jan. 18, 2021, 5:56 a.m.