MakeGSoptions: Encapsulate prior parameters and Gibbs Sampler (GS) control...

In DMRMark: DMR Detection by Non-Homogeneous Hidden Markov Model from Methylation Array Data

Description

This function encapsulate prior parameters and Gibbs Sampler control parameters. All parameters with initial values. The encapsulation is for easy initiating, managing and passing of parameters.

Usage

MakeGSoptions(pi0 = c(100, 100, 5, 5), 
			  cmu0 = c(11.5, 11.5, 8, 8), 
			  theta0 = c(-3, 2), 
			  mu0 = matrix(c(-2, 2, 2, -2), 2, byrow = TRUE), 
			  kappa0 = c(50, 50, 5, 5), 
			  nu0 = rep(4, 2), 
			  A0 = array(rep(c(2, 0.8, 0.8, 4), 2), 
			  dim = c(2, 2, 2)), 
			  alpha12N = rep(40, 3), 
			  beta12N = rep(60, 3), 
			  D_mu = rep(-2, 2), 
			  chi_alpha = 0.2, #This and above for priors 
			  burnin = 500, #This and below for Gibbs Sampler Control
			  nsamples = 100, 
			  sampleSep = 10, 
			  onHMM = TRUE, 
			  track = FALSE, 
			  verbose = FALSE)

Arguments

pi0	Length-4 vector, the concentration of Dirichlet distribution. Prior of initial states.
cmu0	Single value, the mean of Normal distribution. Prior of characteristic length.
theta0	Length-2 vector, each value is the mean of a Normal distributions. Priors for means of control groups of two non-differentially methylated CpG sites (non-DMCs) responses.
mu0	2-by-2 matrix, each row is the means of a bivariate Normal distributions. Priors for means of two DMCs responses
kappa0	Length-4 vector, each value is the prior observation number of Normal-Inverse-Gamma (NIG) or Normal-Inverse-Wishart (NIW) depends on the corresponding state.
nu0	Length-2 vector, each value is the degree of freedom of an IW distribution. Priors for covariance of DMC responses.
A0	2-by-2-by-2 array, each 2-by-2 matrix along the third dimension is the scale matrix of an IW distribution. Priors for covariance of DMC responses.
alpha12N	Length-3 vector, each value is the shape of an IG distribution. Priors for variance of non-DMC responses.
beta12N	Length-3 vector, each value is the rate of an IG distribution. Priors for variance of non-DMC responses.
D_mu	Length-2 vector, each value is the minimum distance between two group means of DMCs. Prior for truncating the means of bivariate normals of DMC's responses.
chi_alpha	p-value of chi-square distribution with 2 degrees of freedom. Prior for truncating the covariant matrices of bivariate normals of DMC's responses.
burnin	Number of iterations for burn-in. Gibbs Sampler control parameter. Default is 500.
nsamples	Number of samples to compute the point estimators. Gibbs Sampler control parameter. Default is 100.
sampleSep	Only keep every 'sampleSep'-th samples to estimate point estimators. Gibbs Sampler control parameter. Default is 10.
onHMM	Set to FALSE will disable HMM, and reduce to simple clustering of Mixture Model. Gibbs Sampler control parameter. Default is TRUE.
track	Set to TRUE will make DMRMark return all samples from the beginning of burn-in to the end of sampling instead of point estimators. Useful for inspecting convergence. Please know well about this issue before you decide to set it to TRUE. Gibbs Sampler control parameter. Default is TRUE.
verbose	Set to TRUE to show the details when running the Gibbs Sampler. Gibbs Sampler control parameter. Default is FALSE.

Value

Simply a list with all items are the same with input. Just an encapsulation.

Author(s)

Linghao SHEN <[e-mail]>

See Also

DMRMark

Examples

	# MakeGSoptions
	opts <- MakeGSoptions()

DMRMark documentation built on May 2, 2019, 1:53 p.m.