用值替换NA-R

Question

Is there a way that I can replace missing values in a vector by randomly sampling from the rest of the data?

eg

age<-c(4.2,5.6,NA,8.4,9.8,NA,10.4,15.3)

age[is.na(age)]<-sample(age,length(age[is.na(age)]),replace=TRUE)  ## trying to replace NA values with a random value from age.

I don't understand why this does not work? Ideally I would like each NA value to be replaced by a different value.

Answer 1

age[is.na(age)] <- sample(age[!is.na(age)], sum(is.na(age)), replace=F)

@Ananda Mahto建议的sum(is.na(age))

Answer 2

In cases where random sampling is undesirable you might replace them with means, medians, or even a model fit:

library(e1071)
?impute
impute(as.matrix(age),what="mean") # replaces with mean 8.95

or

library(randomForest)
?na.roughfix
na.roughfix(age) # replaces with median 9.1

If age is a predictor and you have responses you could use random forest to impute

library(randomForest)
?rfImpute