Usage
score(Y, n.boot=0, score.type="BIC", threshold=0, max.step=500, ini.adj.matrix=NULL,
blacklist=NULL, whitelist=NULL, standardize=TRUE, standardize.boot=TRUE,
random.forest=FALSE, random.step.length=NULL,
nrestart=0, perturb=0, shuffle=FALSE, print=FALSE, EPS=1e-06)
Arguments
Y
an n by p data matrix: n -- sample size, p -- number of variables
n.boot
an integer: the number of bootstrap resamples of the data matrix Y,
default = 0, meaning no bootstrapping
score.type
a string: "BIC" or "likelihood", default = "BIC"
threshold
a nonnegative scalar: the cutoff value for the change of the score
to decide whether to stop the search;
default = 0, meaning stop search when score is not improved
max.step
an integer: the maximum number of search steps of the hill climbing algorithm, default = 500
ini.adj.matrix
a p by p 0-1 matrix: the initial graph, default = NULL, meaning the empty graph
blacklist
a p by p 0-1 matrix: if the (i,j)th-entry is "1",
then the edge i-->j will be excluded from the DAG during the search, default= NULL
whitelist
a p by p 0-1 matrix: if the (i,j)th-entry is "1",
then the edge i-->j will always be included in the DAG during the search, default=NULL
standardize
logic: whether to standardize the data to have mean zero and sd one, default = TRUE
standardize.boot
logic: whether to standardize the bootstrap resamples, default = TRUE
random.forest
logic: whether to use the "random forest" idea for further variance reduction, default=FALSE
random.step.length
a vector: specify ``random forest" steps
nrestart
an integer: number of times to restart the search algorithm after a local optimal is achieved.
The purpose is to search for global optimal, default=0, meaning no restart.
perturb
an integer: how many random addition/deletion/reversal operations
should be used in each random restart,
default = 0, corresponding to no restart.
shuffle
logic: whether to shuffle the order of variables before DAG learning.
The purpose is to avoid potential systematic biases in simulation studies
(due to possible coincidence of the topological ordering and the nominal ordering of the variables), default=FALSE
print
logic: whether print the step information, default= FALSE
EPS
a scalar: a number to indicate what small values will be treated as zero,
default = 1e-06, meaning that values with magnitude smaller than 1e-6 will be treated as zero.