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.