Imputation by Tricube Predictive Mean Matching

This function performs tricube predictive mean matching (see Hmisc::aregImpute) in which donors are weighted according to distances of predicted values. Three donors are chosen.

mice.impute.tricube.pmm(y, ry, x, tricube.pmm.scale=0.2, tricube.boot=FALSE, ...)

Arguments

y	Incomplete data vector of length n
ry	Vector of missing data pattern (FALSE -- missing, TRUE -- observed)
x	Matrix (n x p) of complete covariates.
tricube.pmm.scale	A scaling factor for tricube matching. The default is 0.2.
tricube.boot	A logical indicating whether tricube matching should be performed using a bootstrap sample
...	Further arguments to be passed

Value

A vector of length nmis=sum(!ry) with imputed values.

See also

Hmisc::aregImpute

Examples

if (FALSE) {
#############################################################################
# EXAMPLE 1: Tricube predictive mean matching for nhanes data
#############################################################################

library(mice)
data(nhanes, package="mice")
set.seed(9090)

#*** Model 1: Use default of tricube predictive mean matching
varnames <- colnames(nhanes)
VV <- length(varnames)
method <- rep("tricube.pmm", VV )
names(method) <- varnames
# imputation with mice
imp.mi1 <- mice::mice( nhanes, m=5, maxit=4, method=method )

#*** Model 2: use item-specific imputation methods
iM2 <- method
iM2["bmi"] <- "pmm6"
# use imputation method 'tricube.pmm' for hyp and chl
# select different scale parameters for these variables
tricube.pmm.scale1 <- list( "hyp"=.15, "chl"=.30 )
imp.mi2 <- miceadds::mice.1chain( nhanes, burnin=5, iter=20, Nimp=4,
               method=iM2, tricube.pmm.scale=tricube.pmm.scale1  )
}