imputeNA_cat.R

var.freq <- table(var)
var.freq <- var.freq[rownames(var.freq) != ""]

# impute with mode, most frequent category
var.mode <- names(var.freq[var.freq==max(var.freq)])
var1 <- ifelse(is.na(var) | var=="", var.mode, var)

# impute with random category using sample distribution
var.filled <- var[!is.na(var) & var!=""]
var.random <- var.filled[sample(1:length(var.filled), 1)]
var2 <- ifelse(is.na(var) | var=="", var.random, var)

# impute with random category if actual distr is not representative
var.categories <- rownames(var.freq)
var.random <- var.categories[sample(1:length(var.categories), 1)]
var3 <- ifelse(is.na(var) | var=="", var.random, var)

# impute with its own category
var4 <- ifelse(is.na(var) | var=="", "-1", var)