NormMixClus_K: Normal mixture model estimation

In coseq: Co-Expression Analysis of Sequencing Data

Description

Perform co-expression and co-abudance analysis of high-throughput sequencing data, with or without data transformation, using a Normal mixture models for single number of clusters K. The output of NormMixClus_K is an S3 object of class NormMixClus_K.

Usage

NormMixClus_K(y_profiles, K, alg.type = "EM", init.runs = 50,
  init.type = "small-em", GaussianModel = "Gaussian_pk_Lk_Ck",
  init.iter = 20, iter = 1000, cutoff = 0.001, verbose = TRUE,
  digits = 3)

Arguments

y_profiles	y (n x q) matrix of observed profiles for n observations and q variables
K	Number of clusters (a single value).
alg.type	Algorithm to be used for parameter estimation: “EM”, “CEM”, “SEM”
init.runs	Number of runs to be used for the Small-EM strategy, with a default value of 50
init.type	Type of initialization strategy to be used: “small-em” for the Small-EM strategy, “random”, “CEM”, or “SEMMax”
GaussianModel	One of the 28 forms of Gaussian models defined in Rmixmod, by default equal to the "Gaussian_pk_Lk_Ck" (i.e., a general family model with free proportions, free volume, free shape, and free orientation)
init.iter	Number of iterations to be used within each run for the Small-EM strategry, with a default value of 20
iter	Maximum number of iterations to be run for the chosen algorithm
cutoff	Cutoff to declare algorithm convergence
verbose	If TRUE, verbose output is created
digits	Integer indicating the number of decimal places to be used for the probaPost output

Value

An S3 object of class NormMixClus_K containing the following:

probaPost	Matrix containing the conditional probabilities of belonging to each cluster for all observations
log.like	Value of log likelihood
ICL	Value of ICL criterion
nbCluster	Number of clusters (equivalent to K)
GaussianModel	Gaussian model form fit in the mixture model

Author(s)

Cathy Maugis-Rabusseau, Andrea Rau

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.