trMVN.dat: Transformation of Normal Scores

In MultiVarMI: Multiple Imputation for Multivariate Data

Description

This function backtransforms normal scores for ordinal variables using the thresholds determined by the marginal proportions using quantiles of the normal distribution; normal scores for continuous variables by the sum of linear combinations of standard normals using the corresponding Fleishman coefficients; and normal scores for count variables by the inverse cdf matching procedure.

Usage

trMVN.dat(indat, ord.mps=NULL, nct.sum=NULL, count.rate=NULL)

Arguments

indat	A list of data frames of normal scores to be backtransformed.
ord.mps	A list containing marginal probabilities for binary and ordinal variables as packaged from output in ordmps. Default is NULL.
nct.sum	A matrix containing summary statistics for continuous variables as packaged from output in nctsum. Default is NULL.
count.rate	A vector containing rates for count variables as packaged from output in countrate. Default is NULL.

Value

A list containing backtransformed data.

References

Fleishman A.I. (1978). A method for simulating non-normal distributions. Psychometrika, 43(4), 521-532.

See Also

MI, ordmps, nctsum, countrate

Examples

sndat<-data.frame(matrix(rnorm(1e4), ncol=5, nrow=1e4/5))

#ordinal marginal probabilities
m1<-c(0.4, 0.6)
names(m1)<-c(0,1)
m2<-c(0.2, 0.3, 0.5)
names(m2)<-c(0,2,3)
mps<-list(X1=m1, X2=m2)

#count rates
rates<-c(2, 3)
names(rates)<-c('X3', 'X4')

#continuous
nctsum<-data.frame(X5=c(1, 1, -0.31375, 0.82632, 0.31375, 0.02271)) #Weibull(1,1)
rownames(nctsum)<-c('Mean', 'Variance', 'a', 'b', 'c', 'd')

trdat<-trMVN.dat(indat=list(sndat), ord.mps=mps, nct.sum=nctsum, count.rate=rates)

MultiVarMI documentation built on May 1, 2019, 8:44 p.m.