| Title: | Missing Value Imputation using Random Forest for Prediction Settings |
|---|---|
| Description: | Missing data imputation based on the 'missForest' algorithm (Stekhoven, Daniel J (2012) <doi:10.1093/bioinformatics/btr597>) with adaptations for prediction settings. The function missForest() is used to impute a (training) dataset with missing values and to learn imputation models that can be later used for imputing new observations. The function missForestPredict() is used to impute one or multiple new observations (test set) using the models learned on the training data. |
| Authors: | Elena Albu [aut, cre] |
| Maintainer: | Elena Albu <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 1.0 |
| Built: | 2024-11-07 03:29:26 UTC |
| Source: | https://github.com/sibipx/missforestpredict |
Calculates convergence based on NMSE. Details on the convergence criterion calculation are provided in the package vignettes.
calculate_convergence(err, weights)calculate_convergence(err, weights)
err |
dataframe containing OOB or apparent errors for each iteration. |
weights |
vector of weights in the same format as for the |
A list with elements
converged |
boolean indicating if the algorithm has converged (TRUE) or not (FALSE) |
measure_old |
the total error of the previous iteration |
measure_new |
the total error of the last iteration |
Performs a number of checks on a custom predictor matrix and returns error if the matrix is not in a valid format.
check_predictor_matrix(predictor_matrix, data, verbose = TRUE)check_predictor_matrix(predictor_matrix, data, verbose = TRUE)
predictor_matrix |
custom predictor matrix |
data |
dataframe to be imputed |
verbose |
if TRUE, returns human readable message for the checks performed. |
No return value, only verifies the validity the provided predictor matrix
Creates a square predictor matrix with number of rows and columns equal to the number of columns of the data. Each row name represents the name of the column to be imputed. Each column name represents a predictor for the imputation model. The values in a row will contain 1 for the variables that should be included as predictors in the imputation model and 0 if the variable should not be included as predictor. The diagonal of the predictor matrix is 0, indicating that variable X will not be a predictor for the imputation model of variable X. All other values are 1, meaning that all other variables will be included in the imputation model. This is the default predictor matrix used.
create_predictor_matrix(data)create_predictor_matrix(data)
data |
dataframe to be imputed |
predictor matrix that can be used as a start for setting a custom predictor matrix
Evaluate the imputation error when true values are known. Useful when missing values are simulated and true values are known (the errors are calculated as distances from the true values). For continuous variables, MSE (mean square error) and NMSE (normalized mean square error) are returned. For categorical variables, MER (misclassification error rate) is returned.
evaluate_imputation_error(ximp, xmis, xtrue, all = FALSE)evaluate_imputation_error(ximp, xmis, xtrue, all = FALSE)
ximp |
imputed dataframe. |
xmis |
original dataframe with missing values. |
xtrue |
true dataframe with no missing values. |
all |
calculate error on all observations (TRUE) or only on missing observations (FALSE). Default is FALSE. |
For details check the advanced vignette on convergence criteria and error monitoring.
Dataframe with variables in rows and performance measures in columns.
Imputes a dataframe and (if save_models = TRUE) returns imputation models to be used on new observations.
missForest( xmis, maxiter = 10, fixed_maxiter = FALSE, var_weights = NULL, decreasing = FALSE, initialization = "mean/mode", x_init = NULL, class.weights = NULL, return_integer_as_integer = FALSE, save_models = TRUE, predictor_matrix = NULL, proportion_usable_cases = c(1, 0), verbose = TRUE, convergence_error = "OOB", ... )missForest( xmis, maxiter = 10, fixed_maxiter = FALSE, var_weights = NULL, decreasing = FALSE, initialization = "mean/mode", x_init = NULL, class.weights = NULL, return_integer_as_integer = FALSE, save_models = TRUE, predictor_matrix = NULL, proportion_usable_cases = c(1, 0), verbose = TRUE, convergence_error = "OOB", ... )
xmis |
dataframe containing missing values of class dataframe ("tibble" class tbl_df is also supported). Matrix format is not supported. See details for column format. |
maxiter |
maximum number of iterations. By default the algorithm will stop when converge is reached or after running for maxiter, whichever occurs first. |
fixed_maxiter |
if set to TRUE, the algorithm will run for the exact number of iterations specified in maxiter, regardless of the convergence criteria. Default is FALSE. |
var_weights |
named vector of weights for each variable in the convergence criteria. The names should correspond to variable names. By default the weights are set to the proportion of missing values on each variable. |
decreasing |
(boolean) if TRUE the order in which the variables are imputed is by decreasing amount of missing values. (the variable with highest amount of missing values will be imputed first). If FALSE the variable with lowest amount of missing values will be imputed first. |
initialization |
initialization method before running RF models; supported: mean/mode, median/mode and custom. Default is mean/mode. |
x_init |
if |
class.weights |
a named list containing |
return_integer_as_integer |
Internally, integer columns are treated as double (double precision floating point numbers). If TRUE, the imputations will be rounded to closest integer and returned as integer (This might be desirable for count variables). If FALSE, integer columns will be returned as double (This might be desirable, for example, for patient age imputation). Default is FALSE. The same behaviour will be applied to new observations when using missForestPredict. |
save_models |
if TRUE, imputation models are saved and a new observation (or a test set) can be imputed using the models learned; saving models on a dataset with a high number of variables will occupy RAM memory on the machine. Default is TRUE. |
predictor_matrix |
predictor matrix indicating which variables to use in the imputation of each variable.
See documentation for function |
proportion_usable_cases |
a vector with two components: the first one is a minimum threshold for |
verbose |
(boolean) if TRUE then missForest returns OOB error estimates (MSE and NMSE) and runtime. |
convergence_error |
Which error should be used for the convergence criterion. Supported values: OOB and apparent. If a different value is provided, it defaults to OOB. See vignette for full details on convergence. |
... |
other arguments passed to ranger function (some arguments that are specific to each variable type are not supported).
See vignette for |
An adaptation of the original missForest algorithm (Stekhoven et al. 2012) is used. Variables are initialized with a mean/mode, median/mode or custom imputation. Then, they are imputed iteratively "on the fly" for a maximum number of iterations or until the convergence criteria are met. The imputation sequence is either increasing or decreasing. At each iteration, a random forest model is build for each variable using as outcome on the observed (non-missing) values of the variable and as predictors the values of the other variables from previous iteration for the first variable in the sequence or current iteration for next variables in the sequence (on-the-fly). The ranger package (Wright et al. 2017) is used for building the random forest models.
The convergence criterion is based on the out-of-boostrap (OOB) error or the apparent error and uses NMSE (normalized mean squared error) for both continuous and categorical variables.
Imputation models for all variables and all iterations are saved (if save_models is TRUE) and can be later
applied to new observations.
Both dataframe and tibble (tbl_df class) are supported as input. The imputed dataframe will be retured with the same class. Numeric and integer columns are supported and treated internally as continuous variables. Factor and character columns are supported and treated internally as categorical variables. Other types (like boolean or dates) are not supported. NA values are considered missing values.
Object of class missForest with elements
ximp |
dataframe with imputed values |
init |
x_init if custom initalization is used; otherwise list of mean/mode or median/mode for each variable |
initialization |
value of initialization parameter |
impute_sequence |
vector variable names in the order in which imputation has been run |
maxiter |
maxiter parameter as passed to the function |
models |
list of random forest models for each iteration |
return_integer_as_integer |
Parameter return_integer_as_integer as passed to the function |
integer_columns |
list of columns of integer type in the data |
OOB_err |
dataframe with out-of-bag errors for each iteration and each variable |
Stekhoven, D. J., & Bühlmann, P. (2012). MissForest-non-parametric missing value imputation for mixed-type data. Bioinformatics, 28(1), 112-118. doi:10.1093/bioinformatics/btr597
Wright, M. N. & Ziegler, A. (2017). ranger: A fast implementation of random forests for high dimensional data in C++ and R. J Stat Softw 77:1-17. doi:10.18637/jss.v077.i01.
data(iris) iris_mis <- produce_NA(iris, proportion = 0.1) imputation_object <- missForest(iris_mis, num.threads = 2) iris_imp <- imputation_object$ximpdata(iris) iris_mis <- produce_NA(iris, proportion = 0.1) imputation_object <- missForest(iris_mis, num.threads = 2) iris_imp <- imputation_object$ximp
Imputes a new dataframe based on the missForest models. The same number of iterations as in missForest are used.
missForestPredict(missForestObj, newdata, x_init = NULL)missForestPredict(missForestObj, newdata, x_init = NULL)
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. |
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.
an imputed dataframe
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)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)
Produces a dataframe with missing values.
For each variable, a random sample of observations of size proportion * nrow(x) will be set to NA.
produce_NA(x, proportion = 0.1)produce_NA(x, proportion = 0.1)
x |
a dataframe |
proportion |
proportion of missing values to be produced; a vector of size |
dataframe with missing values.
Calculates variable-wise proportion of usable cases (missing and observed) as in Molenberghs et al. (2014).
prop_usable_cases(data)prop_usable_cases(data)
data |
dataframe to be imputed |
missForest builds models for each variable using the observed values of that variable as outcome of a random forest model. It then imputes the missing part of the variable using the learned models.
If all values of a predictor are missing among the observed value of the outcome,
the value of p_obs will be 1 and the model built will rely heavily on the initialized values.
If all values of a predictor are observed among the observed values of the outcome, p_obs will be 0
and the model will rely on observed values. Low values of p_obs are preferred.
Similarly, if all values of a predictor are missing among the missing values of the outcome,
p_miss will have a value of 0 and the imputations (predictions) will heavily rely on the initialized values.
If all values of a predictor are observed among the missing value of the outcome, p_miss will have a value of 1
and the imputations (predictions) will rely on real values. High values of p_miss are preferred.
Each row represents a variable to be imputed and each column a predictor.
a list with two elements: p_obs and p_miss
p_obs |
the proportion of missing |
p_miss |
the proportion of observed |
Molenberghs, G., Fitzmaurice, G., Kenward, M. G., Tsiatis, A., & Verbeke, G. (Eds.). (2014). Handbook of missing data methodology. CRC Press. Chapter "Multiple Imputation"