impute_missing: Missing values imputation wrapper

In smartdata: Data Preprocessing

Description

Missing values imputation wrapper

Usage

impute_missing(dataset, method, exclude = NULL, ...)

Arguments

dataset	we want to impute missing values on
method	selected method of missing values imputation
exclude	character. Vector of attributes to exclude from the missing values treatment
...	Further arguments for method

Value

The treated dataset (either with noisy instances replaced or erased)

Examples

library("smartdata")
data(africa, package = "Amelia")
data(nhanes, package = "mice")
data(ozone,  package = "missMDA")
data(vnf,    package = "missMDA")
data(orange, package = "missMDA")
data(sleep,  package = "VIM")

super_nhanes <- impute_missing(nhanes, "gibbs_sampling")
super_nhanes <- impute_missing(nhanes, "gibbs_sampling", exclude = "chl")
# Use a different method for every column
impute_methods <- c("pmm", "midastouch", "norm_nob", "norm_boot")
super_nhanes <- impute_missing(nhanes, "gibbs_sampling", imputation = impute_methods)
super_nhanes <- impute_missing(nhanes, "central_imputation")
super_africa <- impute_missing(africa, "knn_imputation")
# Execute knn imputation with non default value for k
super_africa <- impute_missing(africa, "knn_imputation", k = 5)
super_africa <- impute_missing(africa, "expect_maximization", exclude = "country")
super_africa <- impute_missing(africa, "rf_imputation", num_iterations = 15,
                               num_trees = 200, bootstrap = FALSE)
# Examples of calls to 'PCA imputation' with wholly numeric datasets

super_orange <- impute_missing(orange, "PCA_imputation", num_dimensions = 5,
                               imputation = "EM")
super_orange <- impute_missing(orange, "PCA_imputation", num_dimensions = 5,
                               imputation = "Regularized")

super_orange <- impute_missing(orange, "PCA_imputation", num_dimensions = 5,
                               imputation = "Regularized", random_init = TRUE)
# Examples of calls to 'MCA imputation' with wholly categorical datasets

super_vnf    <- impute_missing(vnf, "MCA_imputation", num_dimensions = 5,
                               imputation = "EM")
super_vnf    <- impute_missing(vnf, "MCA_imputation", num_dimensions = 5,
                               imputation = "Regularized")

super_vnf    <- impute_missing(vnf, "MCA_imputation", num_dimensions = 5,
                               imputation = "Regularized", random_init = TRUE)
# Examples of calls to 'FAMD imputation' with hybrid datasets

super_ozone  <- impute_missing(ozone, "FAMD_imputation", num_dimensions = 5,
                               imputation = "EM", exclude = c("Ne12", "Vx15"))
super_ozone  <- impute_missing(ozone, "FAMD_imputation", num_dimensions = 5,
                               imputation = "Regularized")

super_ozone  <- impute_missing(ozone, "FAMD_imputation", num_dimensions = 5,
                               imputation = "Regularized", random_init = TRUE)

# Examples of hotdeck, iterative robust and reggresion imputations
super_sleep <- impute_missing(sleep, "hotdeck")
super_sleep <- impute_missing(sleep, "iterative_robust", initialization = "median",
                              num_iterations = 1000)
super_sleep <- impute_missing(sleep, "regression_imputation",
                              formula = Dream+NonD~BodyWgt+BrainWgt)

# Examples of adaptative shrinkage imputation
super_ozone <- impute_missing(ozone, "ATN", sigma = 2.2)
super_ozone <- impute_missing(ozone, "ATN", lambda = 0.025, gamma = 2.5)
super_ozone <- impute_missing(ozone, "ATN", tune = "SURE")

smartdata documentation built on Dec. 19, 2019, 1:08 a.m.