multimove_gibbs: Draws factor conditional on parameters

In h4sci/packagr: Estimation of a Bayesian Dynamic Factor Model

Description

The Multi-Move Gibbs sampler applies the kalman filter by forward and backwards filtering.

Usage

multimove_gibbs(yt, phi, Q, lambda, const, Tt, q, alpha_0, P_0, R)

Arguments

yt	A matrix of demeaned and standardized time series data.
phi	Diagoanl matrix of dimension k x k with vector autoregressive coefficients.
Q	A matrix of .
lambda	A vector of dimension n x k of the factor loadings.
const	A scalar, where const = 1 for model with intercept, const = 0 for model without intercept.
Tt	Number of high-frequency periods.
q	lag length for state equation (adjust starting value of phi accordingly).
alpha_0	Vector of dimension m x 1 (Initial conditions for Kalman filter).
P_0	Diagonal matrix of dimension m (Initial conditions for Kalman filter).
R	Diagonal matrix of dimension n of idiosyncratic component.

Value

A Vector of factors conditional parameters.

Examples

q <- 1
yt <- as.matrix(t(Xmat))
n <- dim(yt)[1]
Tt <- dim(yt)[2]
k <- 2
m <- k*q
alpha_0 <- matrix(0,m,1)
P_0 <- diag(m)
const <- 0
R <- as.matrix(diag(n)*0.01)
phi <- diag(rnorm(k,0,1))
lambdasim <- matrix(rep(rnorm(n,0,1)*0.1,k),
nrow = n, ncol = k, byrow = TRUE)
diag(lambdasim) <- 1
lambdasim[upper.tri(lambdasim)] <- 0
lambda <- lambdasim
Q <- as.matrix(diag(0.1,k))
matcomp <- comp_f_state(phi,Q,lambda,const)

h4sci/packagr documentation built on Jan. 7, 2021, 10:40 p.m.