zdifference: zdifference for dataset

In weightedZdiff: Calculation of z-Differences

Description

The function calculates the zdifferences for each variable in a dataset or each column in a matrix (depends on the format of your data). Furthermore the sum of the squared zdifferences is calculated. The variables are set into classes continuous, binary and nominal automatically by the following algorithm. If the variable has only 2 different values its treated as binary. If the variable has more then 9 observations or the class of the variable is factor its treated as nominal and otherwise continuous. The user can specify the type of every variable by hand.

Usage

zdifference(dataset,ref,weights=NULL,standard_weights=FALSE,na.rm=TRUE,
binary_variable=NULL,ordinal_variable=NULL,continuous_variable=NULL,nominal_variable=NULL,
r=2,var.est=FALSE,coefvar.est=FALSE,grad=1)

Arguments

dataset	An object of class data.frame or matrix, which contains the variables for which the zDifferences should be calculated and the reference variable in columns.
ref	The name of the reference variable, name must be in datasets' names.
weights	The name of the variable containing the weights for each observation, name must be in datasets' names.
standard_weights	Should the unweighted zdifferences be calculated or not.
na.rm	Should NAs be removed or not. If NAs exists in dataset and na.rm=FALSE then an error will occure.
binary_variable	optional: Names of binomial variables.
ordinal_variable	optional: Names of ordinal variables.
continuous_variable	optional: Names of continuous variables.
nominal_variable	optional: Names of nominal variables.
r	Number of digits to round the result.
var.est	Should the weighted z-Difference for the variances of continuous variables be reported (TRUE) or not (FALSE)
coefvar.est	Should the coefficient of variation for continuous variables be reported (TRUE) or not (FALSE)
grad	The Moments for which to calculate the weighted z-Difference for continuous variables. grad=2 means the first and second moments are calculated.

Author(s)

Tim Filla

References

For standard z-difference (unweighted) https://pubmed.ncbi.nlm.nih.gov/23972521/

Examples

data(testdata)
#new dataset
zdifference(testdata,"treatment",grad=2,continuous_variable=c("age","meanbp1"),
binary_variable=c("CHF","Cirr","colcan","Coma","lungcan","MOSF","sepsis","female","ARF"))
#generate iptw weights
p<-glm(treatment~.,data=testdata,family="binomial")$fitted.values
testdata$weights<-ifelse(testdata$treatment==0,1/(1-p),1/p)
zdifference(testdata,"treatment",weights="weights",grad=2,
continuous_variable=c("age","meanbp1"),binary_variable=c("CHF","Cirr",
"colcan","Coma","lungcan","MOSF","sepsis","female","ARF"),standard_weights=TRUE)

weightedZdiff documentation built on Aug. 18, 2020, 5:07 p.m.