diagnostic_plot: Check the convergence of a data set computed by 'compute_GVA'

In VBel: Variational Bayes for Fast and Accurate Empirical Likelihood Inference

Check the convergence of a data set computed by compute_GVA

Description

Plots mu and variance in a time series plot to check for convergence of the computed data (i.e. Full-Covariance Gaussian VB Empirical Likelihood Posterior)

Usage

diagnostic_plot(dataList, muList, cList)

Arguments

dataList	Named list of data generated from compute_GVA
muList	Array of indices of mu_arr to plot. (default:all)
cList	Matrix of indices of variance to plot, 2xn matrix, each row is (col,row) of variance matrix

Value

Matrix of variance of C_FC

Examples

# -----------------------------
# Initialise Inputs
# -----------------------------
# Generating 30 data points from a simple linear-regression model
seedNum <- 100
set.seed(seedNum)
n       <- 100
p       <- 10
lam0    <- matrix(0, nrow = p)
mean    <- rep(1, p)
sigStar <- matrix(0.4, p, p) + diag(0.6, p)
z       <- mvtnorm::rmvnorm(n = n-1, mean = mean, sigma = sigStar)

# Specify moment condition functions for linear regression and its corresponding derivative
h       <- function(zi, th) { matrix(zi - th, nrow = 10) }
delthh  <- function(z, th) { -diag(p) }

# Specify derivative of log prior
delth_logpi <- function(theta) {-0.0001 * theta}

# Specify AEL constant and Newton-Rhapson iteration
AEL_iters <- 5 # Number of iterations for AEL
a         <- 0.00001

# Specify initial values for GVA mean vector and Cholesky
zbar    <- 1/(n-1) * matrix(colSums(z), nrow = p)
mu_0    <- matrix(zbar, p, 1)

sumVal  <- matrix(0, nrow = p, ncol = p)
for (i in 1:p) {
zi      <- matrix(z[i,], nrow = p)
sumVal  <- sumVal + (zi - zbar) %*% matrix(zi - zbar, ncol = p)
}
sigHat  <- 1/(n-2) * sumVal
C_0     <- 1/sqrt(n) * t(chol(sigHat))

# Specify details for ADADELTA (Stochastic Gradient-Descent)
SDG_iters   <- 5 # Number of iterations for GVA
epsil       <- 10^-5
rho         <- 0.9

# -----------------------------
# Main
# -----------------------------
# Compute GVA
ansGVA <-compute_GVA(mu_0, C_0, h, delthh, delth_logpi, z, lam0, rho, epsil, 
a, SDG_iters, AEL_iters)

# Plot graphs
diagnostic_plot(ansGVA) # Plot all graphs
diagnostic_plot(ansGVA, muList = c(1,4)) # Limit number of graphs
diagnostic_plot(ansGVA, cList = matrix(c(1,1, 5,6, 3,3), ncol = 2)) # Limit number of graphs

VBel documentation built on April 4, 2025, 2:24 a.m.