Engineering missing values (NA) in numerical variables - Random sample imputation (part 2)