makeResiduals: Generate residuals based on variables in imputed data sets

In micd: Multiple Imputation in Causal Graph Discovery

Generate residuals based on variables in imputed data sets

Description

Generate residuals based on variables in imputed data sets

Usage

makeResiduals(data, v, confounder, method = c("res", "cc", "pd"))

Arguments

data	A data.frame.
v	Vector of integers referring to the location of the variable(s) in the data set
confounder	Vector of integers referring to the location of the variable(s) in the data set (confounders are not included in the network!)
method	Default method 'res' uses residuals, 'cc' uses complete cases and 'pd' uses pairwise deletion

Value

A data matrix of residuals.

Examples

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

# Impute missing values
imp <- mice(daten, m = 5)

# Build residuals
knoten <- 1:4
confounder <- 5:6

# Residuals based on dataset with missing values
res.pd <- makeResiduals(daten, v = knoten, confounder = confounder, method = "pd")

# Residuals based in multiple imputed data
residuals <- list(data = list(), m = 5)
imp_c <- mice::complete(imp, "all")
for (i in 1:imp$m){
   residuals$data[[i]] <- makeResiduals(imp_c[[i]],
                          v = knoten, confounder = confounder)
 }

pc.res <- pcMI(data = residuals, p = length(knoten), alpha = 0.05)
fci.res <- fciMI(data = imp, p = length(knoten), alpha = 0.05)

if (requireNamespace("Rgraphviz", quietly = TRUE)){
oldpar <- par(mfrow = c(1,2)) 
  plot(pc.res)
  plot(fci.res)
par(oldpar)
}

micd documentation built on March 7, 2023, 6:03 p.m.