GMM_kmeans: Initialize parameters for modified GMM using kmeans...

In hillarykoch/CLIMB: Composite LIkelihood eMpirical Bayes

Initialize parameters for modified GMM using kmeans clustering.

Description

For a given number of clusters, find initial clusters and compute initial parameter estimates.

Usage

GMM_kmeans(y, k, iter.max = 30)

Arguments

y	An n by d numeric matrix with n observations of dimension d.
k	Number of cluster centers, passed to kmeans.
iter.max	The maximum number of iterations allowed, passed to kmeans.

Value

prop	Proportion of observations assigned to each cluster.
mu	Vector of cluster means.
sigma	List of cluster variance covariance matrices.
cluster	Vecor of cluster labels.

Author(s)

[e-mail]

Examples

sim <- rGMCM(1500, prop = c(1/3,1/3,1/3),
            mu = list(c(-1,1), c(1,1), c(0,-sqrt(2))),
            sigma = list(rbind(c(0.65, 0.7794), c(0.7794, 1.55)),
                        rbind(c(0.65, -0.7794), c(-0.7794, 1.55)),
                        diag(c(2, 0.2))))
init <- GMM_kmeans(sim$data, k = 3)

hillarykoch/CLIMB documentation built on Oct. 24, 2022, 4:27 a.m.