| gibbs_all {CNPBayes} | R Documentation |
Evaluate both single-batch and multi-batch models with the specified range for the number of components, returning the top models sorted by marginal likelihood
gibbs_all(hp.list, mp, dat, batches, k_range = c(1, 4), max_burnin = 32000, top = 3)
hp.list |
a list of hyperparameters. See example. |
mp |
a |
dat |
numeric vector of CNP summary statistics (e.g., median log R ratios) |
batches |
an integer vector of the same length as the data providing an index for the batch |
k_range |
a length-two integer vector providing the minimum and maximum number of components |
max_burnin |
a length-one integer vector indicating the maximum number of burnin iterations |
top |
the number of models to return after ordering by the marginal likelihood |
a list of models
set.seed(100)
nbatch <- 3
k <- 3
means <- matrix(c(-2.1, -2, -1.95, -0.41, -0.4, -0.395, -0.1,
0, 0.05), nbatch, k, byrow = FALSE)
sds <- matrix(0.15, nbatch, k)
sds[, 1] <- 0.3
N <- 1000
truth <- simulateBatchData(N = N, batch = rep(letters[1:3],
length.out = N),
p = c(1/10, 1/5, 1 - 0.1 - 0.2),
theta = means,
sds = sds)
hp <- HyperparametersMultiBatch(k=3,
mu=-0.75,
tau2.0=0.4,
eta.0=32,
m2.0=0.5)
hp.sb <- Hyperparameters(tau2.0=0.4,
mu.0=-0.75,
eta.0=32,
m2.0=0.5)
hp.list <- list(single_batch=hp.sb,
multi_batch=hp)
mp <- McmcParams(iter = 1000,
burnin = 1000,
nStarts = 4,
thin=10)
## Not run:
models <- gibbs_all(hp.list=hp.list, dat=y(truth),
batches=batch(truth),
mp=mp,
top=3)
## End(Not run)