r_truncated_multivariate_normal: Simulate from a truncated multivariate normal distribution

In bomarkussen/probit: Item response analysis with random effects.

Description

Simulate S in R^M and Z in R^d given S in (alpha,beta]$, where S = X + gamma^T Z with X ~ Normal(0,I_N) and Z ~ Normal(0,Gamma^T Gamma).

Usage

r_truncated_multivariate_normal(
  n = 1,
  alpha,
  beta,
  gamma,
  Gamma,
  which = "Z",
  eps = 1e-12,
  maxit = 100
)

Arguments

n	Number of samples. Default: n=1.
alpha	Vector of lower bounds for S component.
beta	Vector of upper bounds for S component.
gamma	Matrix of linear transformations on Z component.
Gamma	Cholesky factor for variance on Z component.
which	Text string ("S"/"Z"/"SZ"/"all") deciding which componts are return. Default: which="Z".
eps	Convergence criteria for coupling from the past. Default: eps=1e-12.
maxit	Maximal number of steps in Gibbs sampler. Default: maxit=100.

Value

Simulations either as a matrix (if which="S" or "Z"), or as a list of matrices (if which="SZ").

Examples

r_truncated_multivariate_normal(n=10,alpha=rep(-1,5),beta=rep(1,5),gamma=matrix(rnorm(10),2,5),Gamma=matrix(c(1,1,0,1),2,2))

bomarkussen/probit documentation built on April 3, 2021, 7:38 p.m.