makemar function

Creates artificial missing at random missingness

Introduces missingness into x1 and x2 into a data.frame of the format produced by simdata, for use in the simulation study. The probability of missingness depends on the logistic of the fully observed variables y and x3; hence it is missing at random but not missing completely at random.

makemar(simdata, prop = 0.2)

Arguments

• simdata: simulated dataset created by simdata.
• prop: proportion of missing values to be introduced in x1 and x2.

Details

This function is used for simulation and testing.

Returns

A data.frame with columns:

• y: dependent variable, based on the model y = x1 + x2 + x3 + normal error

• x1: partially observed continuous variable

• x2: partially observed continuous or binary (factor) variable

• x3: fully observed continuous variable

• x4: variable not in the model to predict y, but associated with x1, x2 and x3; used as an auxiliary variable in imputation

See Also

simdata

Examples

set.seed(1)
mydata <- simdata(n=100)
mymardata <- makemar(mydata, prop=0.1)
# Count the number of missing values
sapply(mymardata, function(x){sum(is.na(x))})
# y x1 x2 x3 x4 
# 0 11 10  0  0