MultiScale.dp.kernel: Fit Two-Class Model with Gene-Level Probabilities using...

In lockEF/BayesianScreening: Perform Bayesian Screening with Gene Dependence

Description

Estimates posterior probability of a difference between two sample groups for each genomic variable (e.g., methylation site), using shared kernels and gene-dependent prior probabilities of equality.

Usage

MultiScale.dp.kernel(X,Class,UniqGene,Gene,mu,Sigma,Concentration=0.5,alpha=1,KK=10,NumDraws=1000)

Arguments

X	A matrix in which rows represent variables (e.g., methylation sites) and columns represent samples. The entries of the matrix must be continuous between 0 and 1.
Class	A vector giving a class label for each sample.
UniqGene	Vector giving the gene labels.
Gene	Vector giving the gene label for each variable (length must be equal to the number of rows in X).
mu	Vector of kernel means.
Sigma	Vector of kernel standard deviations
Concentration	Dirichlet hyperparameter for kernel weights
alpha	Dirichlet process concentration parameter for the gene-level probabilities
KK	Stick-breaking threshold for Dirichlet process
NumDraws	Number of MCMC draws for posterior inference.

Value

Returns an object with the following values, averaged over the MCMC iterations:

pG	Vector giving the gene-level prior for association for each gene
posts	Vector giving the posterior probability of association for each marker

Author(s)

Eric F. Lock

References

Lock, E. F. & Dunson, D. B. (2016). Bayesian genome- and epigenome-wide association studies with gene level dependence. Preprint.

Lock, E. F. & Dunson, D. B. (2015). Shared kernel Bayesian screening. Biometrika, 102 (4): 829-842.

lockEF/BayesianScreening documentation built on May 24, 2020, 11:50 p.m.