boot.graph: Bootstrap Resampling for the PC-MI- and the FCI-MI-algorithm

In bips-hb/micd: Multiple Imputation in Causal Graph Discovery

Bootstrap Resampling for the PC-MI- and the FCI-MI-algorithm

Description

Generate R bootstrap replicates for the PC or FCI algorithm for data with missing values.

Usage

boot.graph(
  data,
  select = NULL,
  method = c("pcMI", "fciMI"),
  method.mice = NULL,
  args,
  R,
  m = 10,
  args.residuals = NULL,
  seed = NA,
  quickpred = FALSE,
  ...
)

Arguments

data	Data.frame with missing values
select	Variable of integers, indicating columns to select from a data frame; only continuous variables can be included in the model selection
method	Character string specifying the algorithm for causal discovery from the package 'pcalg'.
method.mice	Character string specifying imputation method; see mice::mice() for more information.
args	Arguments passed to method. NOTE: argument labels is set internally and should not be used!
R	A positive integer number of bootstrap replications.
m	Number of chains included in mice()'.
args.residuals	(Optional) list containing vertices and confounders. May be specified when residuals for vertices should be calculated in each bootstrap data set. See makeResiduals() for more information
seed	A positive integer that is used as argument for set.seed().
quickpred	If true, mice uses quickpred to select predictors.
...	Further arguments passed to the imputation function mice().

Value

List of objects of class pcalgo (see pcalg::pcAlgo) or of fcmialgo (see pcalg::fciAlgo).

Examples

data(windspeed)
daten <- mice::ampute(windspeed)$amp


bgraph <- boot.graph(data = daten,
                     method = "pcMI",
                     args = "solve.confl = TRUE, alpha = 0.05",
                     R = 5)

bips-hb/micd documentation built on Feb. 22, 2023, 7:59 p.m.