missForestPredict {missForestPredict} | R Documentation |
Imputes a new dataframe based on the missForest models
Description
Imputes a new dataframe based on the missForest models. The same number of iterations as in missForest are used.
Usage
missForestPredict(missForestObj, newdata, x_init = NULL)
Arguments
missForestObj |
missForest object as returned by the missForest function. |
newdata |
new data to impute. The column names should be the same as in the imputation model. |
x_init |
initialization dataframe in case custom initialization mode has been used. It needs to be complete dataframe (with no missing values). See vignette for a full example. |
Details
A new observation is initialized in the same manner as passed through the initialization
parameter to the missForest
function. Then, variables are imputed in the same sequence and for the same
number of iterations using the random models saved for each iteration. This ensures that a new observation is
imputed in the same manner as the training set (imputed by the function missForest
).
Re-imputing the training set with the missForestPredict
will yield the same result as
the original imputation returned by the missForest
function.
Value
an imputed dataframe
Examples
data(iris)
# split train / test and create missing values
id_test <- sample(1:nrow(iris), floor(nrow(iris)/3))
iris_train <- iris[-id_test,]
iris_test <- iris[id_test,]
iris_train_miss <- produce_NA(iris_train, proportion = 0.1)
iris_test_miss <- produce_NA(iris_test, proportion = 0.1)
# impute train and learn imputation models
iris_train_imp_obj <- missForest(iris_train_miss, save_models = TRUE, num.threads = 2)
# impute test
iris_test_imp_new <- missForestPredict(iris_train_imp_obj, newdata = iris_test_miss)
head(iris_test_imp_new)