Description Usage Arguments Details Value Author(s) Examples
BayesMP
1 2 3 4 5 6 | BayesMP(Z, gamma = NULL, updateGamma = TRUE, beta = 1/2, alpha = 1,
mu0 = 0, sigma0 = 10, sigma = 1, trunc = 0, empMu = rep(0, ncol(Z)),
empSD = rep(1, ncol(Z)), niter = 100, burnin = 50, silence = FALSE,
logDotsPerLine = 50, fileName = "BayesMP_", writeY = TRUE,
writePi = FALSE, writeDelta = FALSE, writeGamma = FALSE,
writeHSall = TRUE)
|
Z |
Z statistics. Z should be a p*n matrix. |
gamma |
initial gamma Estimated null proportation, by default, gamma will be estimated by the emperical null method. |
updateGamma |
If TRUE, will update gamma by MH method. If FALSE, will keep the gamma as constant. |
beta |
Non-informative prior: given a gene is DE, the prior probablity this gene is up-regulated, default=1/2. |
alpha |
Concentration parameter for DPs, default=1. |
mu0 |
Mean parameter for base function, default=0. |
sigma0 |
sqrt root of variance parameter for base function, default=10. |
sigma |
sqrt root of variance parameter for DP mixture component, default=1. |
trunc |
truncation parameter for base function (For both positive component and negative component), default=0. |
empMu |
a vector of mean parameter for the null component, default is 0. Alternatively, this vector can by estimated by the emperical null method. |
empSD |
a vector of sd parameter for the null component, default is 1. Alternatively, this vector can by estimated by the emperical null method. |
niter |
Number of iterations. Default 100, suggest to be 10,000 |
burnin |
Number of burnin period. Default 50, suggest to be 500 |
silence |
If FALSE (default), will print the MCMC progress in the console. |
logDotsPerLine |
Number of dots printed perline in the console when silence is FALSE. |
fileName |
Base fileName for saving fulll mcmc results. |
writeY |
If TRUE, will save all (niter) posterior samples of Y from MCMC. |
writePi |
If TRUE, will save all (niter) posterior samples of Pi from MCMC. |
writeDelta |
If TRUE, will save all (niter) posterior samples of Delta from MCMC. |
writeGamma |
If TRUE, will save all (niter) posterior samples of Gamma from MCMC. |
writeHSall |
If TRUE, will save the HSall matrix (niter - burnin). Each row of HSall represent a input gene (feature). If the number in the ith column and jth row equals m, it represents there are m posterior samples of Y for feature j that are DE in at least i studies. This matrix will be the input matrix to calculate the Bayesian FDR. For example, for each column i, after normalized (divided) by total number effective samples (niter - burnin), it is the Bayesian belief that the genes are significiant in at least i studies. The Bayesian FDR can be calculated by the BayesianFDR function. |
implementation for BayesMP, MCMC part. This is a full Bayesian model, and the alternative distribution is modeled via dirichlet process.
The MCMC object for the last iteration.
Zhiguang Huo <zhuo@ufl.edu>
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 | set.seed(15213)
G <- 2000
K <- 10
alpha <- 200
X0 <- matrix(rnorm(G * K), G, K)
Xplus <- matrix(rnorm(G * K, 2), G, K)
Xminus <- matrix(rnorm(G * K, -2), G, K)
piall <- rbeta(G, alpha/G, 1)
delta <- rbeta(G, 1/2, 1/2)
p0 <- 1 - piall
p1 <- piall * delta
p2 <- piall * (1 - delta)
Y <- replicate(K, apply(cbind(p0, p1, p2),1,function(x) sample(c(0,1,-1),1,prob = x)))
Z <- X0 * (Y == 0) + Xplus * (Y == 1) + Xminus * (Y == -1)
system.time(BayesMP(Z,writeHSall=F))
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.