ICMEM: ICMEM.

In SC.MEB: Spatial Clustering with Hidden Markov Random Field using Empirical Bayes

Description

The function ICMEM was used to conduct spatial clustering with hidden Markov random field for a sequence of beta and fixed number of clusters

Usage

ICMEM(
  y,
  x_int,
  Adj,
  mu_int,
  sigma_int,
  alpha,
  beta_grid,
  PX,
  maxIter_ICM,
  maxIter
)

Arguments

y	is a matrix of PCs containing gene expression.
x_int	is a vector of initial cluster label.
Adj	is a matrix containing neighborhood information generated by find_neighbors2.
mu_int	is a initial mean vector. we often generated it by Gaussian mixture model.
sigma_int	is a initial co-variance matrix. we often generated it by Gaussian mixture model.
alpha	is a intercept.
beta_grid	is a sequence of smoothing parameter that can be specified by user.
PX	is a logical value specifying the parameter expansion in EM algorithm.
maxIter_ICM	is the maximum iteration of ICM algorithm.
maxIter	is the maximum iteration of EM algorithm.

Details

The function ICMEM was used to conduct spatial clustering with hidden Markov random field for fixed beta and fixed number of clusters

Value

a list.

The item 'x' is the clustering result.

The item 'gam' is the posterior probability matrix.

The item 'ell' is the opposite log-likelihood.

The item 'mu' is the mean of each component.

The item 'sigma' is the variance of each component.

Examples

y = matrix(rnorm(50, 0, 1), 25,2)
pos = cbind(rep(1:5, each=5), rep(1:5, 5))
Adj = getneighborhood_fast(pos, 1.2)
beta_grid = c(0.5,1)
G = 2
fit_int = Mclust(y, G = G)
x_gmm <- fit_int$classification
mu_int <- unname(fit_int$parameter$mean)
sigma_int <- unname(fit_int$parameter$variance$sigma) 
alpha <- -log(fit_int$parameter$pro)*0
reslist <- ICMEM(y = y, x_int = x_gmm, Adj = Adj, mu_int = mu_int, sigma_int = sigma_int,
alpha = alpha, beta_grid = beta_grid, 
PX = TRUE, maxIter_ICM = 10, maxIter = 50) 

SC.MEB documentation built on Oct. 8, 2021, 9:08 a.m.