The function computes (and by default plots) estimates of the autocovariance or autocorrelation function for the different parameters of the model. This is a wrapper for coda::acf.
acfParams(
res,
params = c("w", "xi", "xi0", "psi", "G", "E", "eta"),
only_non_trivial_clusters = TRUE,
lag.max = NULL,
type = c("correlation", "covariance", "partial"),
plot = TRUE,
...
)An acfParamsCOMIX object which is a named list,
with a named element for each requested parameter. Each element is
an object of class acf (from the coda package).
#' @examples
library(COMIX)
# Number of observations for each sample (row) and cluster (column):
njk <-
matrix(
c(
150, 300,
250, 200
),
nrow = 2,
byrow = TRUE
)
# Dimension of data: p <- 3
# Scale and skew parameters for first cluster: Sigma1 <- matrix(0.5, nrow = p, ncol = p) + diag(0.5, nrow = p) alpha1 <- rep(0, p) alpha1[1] <- -5 # location parameter for first cluster in first sample: xi11 <- rep(0, p) # location parameter for first cluster in second sample (aligned with first): xi21 <- rep(0, p)
# Scale and skew parameters for second cluster: Sigma2 <- matrix(-1/3, nrow = p, ncol = p) + diag(1 + 1/3, nrow = p) alpha2 <- rep(0, p) alpha2[2] <- 5 # location parameter for second cluster in first sample: xi12 <- rep(3, p) # location parameter for second cluster in second sample (misaligned with first): xi22 <- rep(4, p)
# Sample data: set.seed(1) Y <- rbind( sn::rmsn(njk[1, 1], xi = xi11, Omega = Sigma1, alpha = alpha1), sn::rmsn(njk[1, 2], xi = xi12, Omega = Sigma2, alpha = alpha2), sn::rmsn(njk[2, 1], xi = xi21, Omega = Sigma1, alpha = alpha1), sn::rmsn(njk[2, 2], xi = xi22, Omega = Sigma2, alpha = alpha2) )
C <- c(rep(1, rowSums(njk)[1]), rep(2, rowSums(njk)[2]))
prior <- list(zeta = 1, K = 10) pmc <- list(naprt = 5, nburn = 200, nsave = 200) # Reasonable usage pmc <- list(naprt = 5, nburn = 2, nsave = 5) # Minimal usage for documentation # Fit the model: res <- comix(Y, C, pmc = pmc, prior = prior)
# Relabel to resolve potential label switching issues: res_relab <- relabelChain(res) effssz <- effectiveSampleSize(res_relab, "w") # Or: tidy_chain <- tidyChain(res_relab, "w") acf_w <- acfParams(tidy_chain, "w")
# (see vignette for a more detailed example)
An object of class COMIX or tidyChainCOMIX.
A character vector naming the parameters to compute and plot the autocorrelation plots for.
Logical, if TRUE only compute and/or
plot the autocorrelation for the clusters that are estimated to be non-empty.
maximum lag at which to calculate the autocorrelation. See more details at ?acf.
Character string giving the type of autocorrelation to be computed. See more details at ?acf.
Logical. If TRUE (the default) the autocorrelation is
plotted.
Other arguments passed to acf.