sample_Wt: Evolution error variance sampler

In drkowal/FDLM: A Bayesian Multivariate Functional Dynamic Linear Model

Description

Sample the (non-dynamic) evolution error variance matrix.

Usage

sample_Wt(resBeta, Rinv = diag(1, K), rh0 = K, useDiagonal = FALSE)

Arguments

resBeta	(T-1) x K matrix of residuals from the evolution equation
Rinv	K x K; only needed for useDiagonal = FALSE
rh0	scalar prior scale parameter; only needed for useDiagonal = FALSE
useDiagonal	logical; when TRUE, use a diagonal covariance matrix; otherwise, use a full matrix

Value

The K x K evolution error covariance matrix, Wt.

Note

The sampler has two options: useDiagonal = TRUE, in which the evolution error variance matrix is diagional with independent Gamma(0.001,0.001) priors on the K diagonal component precisions; and useDiagonal = FALSE, in which the K x K evolution error variance matrix is full with an inverse Wishart prior with prior precision Rinv and scale parameter rh0. The default is Rinv = diag(1, K) and rh0 = K.

Examples

# Simulate an example:
n = 100; K = 3
resBeta = matrix(rnorm(n*K), nr = n, nc = K)

# Sample the variance:
sample_Wt(resBeta)

drkowal/FDLM documentation built on May 20, 2019, 5:20 p.m.