Simulate Covariance Matrices from Prior

Simulates covariance matrices from prior as specified in a MCMCglmm prior

Usage

rprior(n, prior, vtype="us")

Arguments

n: number of observations
prior: list: with elements V, nu and (optionally) alpha.mu and alpha.V
vtype: character: variance structure type with default us

Value

numeric

Details

If alpha.V is NULL the distribution is an inverse-Wishart distribution (i.e. inverse-gamma in the univariate case). If alpha.V is non-null, MCMCglmm uses parameter expansion and the distribution is an inverse-Wishart mixture with no closed form density function expect in the univariate case (scaled F with 1 numerator degree of freedom).

Author

Jarrod Hadfield j.hadfield@ed.ac.uk

Examples


prior<-resolve_prior(F(10, 20), k=3, vtype="us")
# parameter expanded prior for 3x3 covariance matrix with scaled (20) central F_{1,10} marginal variances

V<-rprior(2000, prior)

hist(V[,1], freq=FALSE, breaks=100, main="", xlab="Variance")
x<-seq(1e-6, max(V[,1]), length=1000)

mprior<-resolve_prior(F(10, 20), k=1, vtype="us")
# univariate prior for central F_{1,10} with scale 20

lines(dprior(x, mprior)~x)

# Density for variance