boot.fci: Fast Causal Inference Algorithm (FCI) With Bootstrapping

In rlebron-bioinfo/gnlearn: Genetic Network Learning

Description

This function allows you to learn a directed graph from a dataset using the Fast Causal Inference algorithm (stable version).

Usage

boot.fci(
  df,
  whitelist = NULL,
  blacklist = NULL,
  indep.test = pcalg::gaussCItest,
  alpha = 0.01,
  max.sx = Inf,
  pdsep.max = Inf,
  conservative = FALSE,
  maj.rule = FALSE,
  version = c("fci", "rfci", "fci.plus"),
  type = c("normal", "anytime", "adaptive"),
  rules = rep(TRUE, 10),
  doPdsep = TRUE,
  biCC = FALSE,
  R = 200,
  m = NULL,
  threshold = 0.5,
  to = c("igraph", "adjacency", "edges", "graph", "bnlearn"),
  cluster = parallel::detectCores(),
  seed = sample(1:10^6, 1)
)

Arguments

df	Dataset.
whitelist	A data frame with two columns, containing a set of arcs to be included in the graph (optional).
blacklist	A data frame with two columns, containing a set of arcs not to be included in the graph (optional).
indep.test	Conditional independence test to be used (pcalg implementation). Default: pcalg::gaussCItest
alpha	Target nominal type I error rate. Default: 0.01
max.sx	Maximum allowed size of the conditioning sets.
pdsep.max	Maximum size of Possible-D-SEP for which subsets are considered as conditioning sets in the conditional independence tests.
conservative	Whether or not the conservative PC is used (pcalg implementation). Default: FALSE
maj.rule	Whether or not the triples shall be checked for ambiguity using a majority rule idea, which is less strict than the conservative PC algorithm (pcalg implementation). Default: FALSE
version	Version of FCI algorithm to be used: 'fci', 'rfci', or 'fci.plus'. Default: 'fci'
type	Type of FCI algorithm to be used: 'normal', 'anytime', or 'adaptive'. Default: 'normal'
rules	Logical vector of length 10 indicating which rules should be used when directing edges. Default: rep(TRUE,10)
doPdsep	If FALSE, Possible-D-SEP is not computed, so that the algorithm simplifies to the Modified PC algorithm of Spirtes, Glymour and Scheines (2000, p.84). Default: TRUE
biCC	If TRUE, only nodes on paths between nodes x and y are considered to be in Possible-D-SEP(x) when testing independence between x and y. Default: TRUE
R	Number of bootstrap replicates (optional). Default: 200
m	Size of training set (optional). Default: nrow(df)/2
threshold	Minimum strength required for a coefficient to be included in the average adjacency matrix (optional). Default: 0.5
to	Output format ('adjacency', 'edges', 'graph', 'igraph', or 'bnlearn') (optional).
cluster	A cluster object from package parallel or the number of cores to be used (optional). Default: parallel::detectCores()
seed	Seed used for random selection. Default: NULL

Examples

obj <- boot.fci(df)
avg.g <- obj$average
g.rep <- obj$replicates

rlebron-bioinfo/gnlearn documentation built on July 25, 2020, 12:38 p.m.