CombineMIs: Combine Multiple Imputations

In MGMM: Missingness-Aware Gaussian Mixture Models

Combine Multiple Imputations

Description

Combines point estimates and their estimated sampling (co)variances across multiple imputations using the usual multiple-imputation combining rules.

Usage

CombineMIs(points, covs)

Arguments

points	List of point estimates (each may be a vector or scalar).
covs	List of estimated sampling covariance matrices (or variances for scalar estimates), one per imputation.

Value

List containing the combined point estimate (point) and the combined sampling covariance (cov).

Examples

set.seed(100)

# Generate data and introduce missingness.
data <- rGMM(n = 25, d = 2, k = 1)
data[1, 1] <- NA
data[2, 2] <- NA
data[3, ] <- NA 

# Fit GMM.
fit <- FitGMM(data)

# Lists to store summary statistics.
points <- list()
covs <- list()

# Perform 50 multiple imputations.
# For each, calculate the marginal mean and its sampling variance.
for (i in seq_len(50)) {
  imputed <- GenImputation(fit)
  points[[i]] <- apply(imputed, 2, mean)
  covs[[i]] <- cov(imputed) / nrow(imputed)
}

# Combine summary statistics across imputations.
results <- CombineMIs(points, covs)

MGMM documentation built on Feb. 27, 2026, 1:07 a.m.