fit.em {metaArray}R Documentation

Probability of expresison from mixture distribution for a single gene.

Description

This core function fits two-component normal-uniform mixture distribution, and extracts probability of over/under expression for all samples in all genes.

Usage

  fit.em(x, cl = NULL, threshold=1e-04)

Arguments

x A numeric vector, especially expression values for a particular gene.
cl A vector of 0s and 1s. Use 0 for normal phenotype and 1 for non-normal phenotype. Note that this is the opposite of POE MCMC. Labelling is the opposite in poe.mcmc (will soon be corrected). If left blank, all samples will be labeled as normal phenotype. When class information is provided, conditional estimation of the mixture is applied and POE is estimated unconditionally.
threshold Criterion for convergence in likelihood

Value

expr Estimated POE
a Minimum (adjusted) of Raw Expression
b Maximum (adjusted) of Raw Expression
sigmasq Estimated variance of normal component
mu Estimated mean of normal component
pi Probability that the gene is over/under expressed on average across the samples
conv Convergence status of EM
dec Detection of decreasing likelihood
loc Signs of local maxima
lik.rec Trajectory of likelihood during EM
abnormal Overall sign of failure in EM by any of the three criteria above

Author(s)

Debashis Ghosh <ghoshd@umich.edu>, Hyungwon Choi <hwchoi@umich.edu>

Examples

  

[Package metaArray version 1.4.4 Index]