NormMixParam: Calculate the mean and covariance for a Normal mixture model

In coseq: Co-Expression Analysis of Sequencing Data

Description

Calculates the mean and covariance parameters for a normal mixture model of the form pK_Lk_Ck

Usage

NormMixParam(x, y_profiles = NULL, K = NULL, digits = 3, plot = FALSE,
  ...)

Arguments

x	Object of class coseq, NormMixClus, or NormMixClus_K
y_profiles	y (n x q) matrix of observed profiles for n observations and q variables, required for x of class NormMixClus or NormMixClus_K
K	The model used for parameter estimation for objects x of class coseq or NormMixClus. When NULL, the model selected by the ICL criterion is used; otherwise, K should designate the number of clusters in the desired model
digits	Integer indicating the number of decimal places to be used for output
plot	If true, produce heatmaps to visualize the estimated per-cluster correlation matrices
...	Additional optional parameters to pass to corrplot, if desired

Value

pi	Vector of dimension K with the estimated cluster proportions from the Gaussian mixture model, where K is the number of clusters
mu	Matrix of dimension K x d containing the estimated mean vector from the Gaussian mixture model, where d is the number of samples in the data y_profiles and K is the number of clusters
Sigma	Array of dimension d x d x K containing the estimated covariance matrices from the Gaussian mixture model, where d is the number of samples in the data y_profiles and K is the number of clusters
rho	Array of dimension d x d x K containing the estimated correlation matrices from the Gaussian mixture model, where d is the number of samples in the data y_profiles and K is the number of clusters

Author(s)

Andrea Rau, Cathy Maugis-Rabusseau

Examples

## Simulate toy data, n = 300 observations
set.seed(12345)
countmat <- matrix(runif(300*4, min=0, max=500), nrow=300, ncol=4)
countmat <- countmat[which(rowSums(countmat) > 0),]
profiles <- transform_RNAseq(countmat, norm="none", 
                             transformation="arcsin")$tcounts

conds <- rep(c("A","B","C","D"), each=2)

## Run the Normal mixture model for K = 2,3
run <- NormMixClus(y=profiles, K=2:3, iter=5)

## Run the Normal mixture model for K=2
run2 <- NormMixClus_K(y=profiles, K=2, iter=5)

## Re-estimate mixture parameters for the model with K=2 clusters
param <- NormMixParam(run2, y_profiles=profiles)

## Summary of results
summary(run, y_profiles=profiles)

coseq documentation built on May 2, 2019, 4:55 p.m.