API reference

Everything listed here can be imported directly from the top-level package:

from imputation_methods import MeanImputer, KNNImputer, rmse

The submodules below only group related imputers; you don't need them for imports.

The imputer interface

Every imputer subclasses BaseImputer and exposes one method, impute(df) -> DataFrame:

• Input: a pandas.DataFrame with numeric columns, where missing values are NaN (or pd.NA in nullable dtypes). Non-numeric columns raise TypeError. GroupMeanImputer is the exception: its grouping column may be non-numeric.
• Output: a new dataframe with the same index and columns. The input is never modified. IndicatorImputer also appends one indicator column per input column.
• Output types follow one policy for every imputer:
• a column without missing values comes back unchanged, with the same values and dtype;
• an imputed column comes back as floating point: float32 and float64 keep their precision, and pandas nullable columns (Int64, Float64, ...) become Float64, so values that couldn't be imputed stay <NA>.
• Configuration happens in the constructor. Imputers with randomness accept a random_state for reproducible results.
• Columns with no observed values carry no information, so every imputer that learns from the data leaves them as NaN. Only imputers that fill in a constant you choose, ConstantImputer and IndicatorImputer(strategy="zero"), fill them. To use a fixed value for empty columns, fill them explicitly, for example with HybridImputer(methods=[..., ConstantImputer(0)]).
• Errors: if a model can't be fitted (for example a singular matrix), imputers with an on_error parameter either raise ImputationError (on_error="raise") or fall back to mean or median imputation (on_error="fallback"). Leaving it unset falls back with a FutureWarning; from 1.0.0 the default is "raise".
• Row order matters only for the time-series imputers; sort your data first.

Each imputer also has a functional shortcut, for example knn_impute(df, n_neighbors=3) for KNNImputer(n_neighbors=3).impute(df).

Modules

Page	Module	Contents
Base class	imputation_methods.base	BaseImputer
Statistical	imputation_methods.statistical	MeanImputer, MedianImputer, ModeImputer, ConstantImputer, QuantileImputer, TrimmedMeanImputer, EndOfDistributionImputer, GroupMeanImputer, IndicatorImputer
Donor sampling	imputation_methods.sampling	RandomSamplingImputer, HotDeckImputer, ColdDeckImputer
Time series	imputation_methods.time_series	LOCFImputer, NOCBImputer, ForwardFillFallbackImputer, InterpolationImputer, MovingAverageImputer, WeightedMovingAverageImputer, LinearTrendImputer, PolynomialTrendImputer, SeasonalImputer, KalmanFilterImputer
Nearest neighbors	imputation_methods.neighbors	KNNImputer, RadiusNeighborsImputer, LocalMeanImputer
Regression	imputation_methods.regression	RegressionImputer, StochasticRegressionImputer, PMMImputer, BayesianRidgeImputer, HuberImputer, RANSACImputer, GaussianProcessImputer
Iterative	imputation_methods.iterative	MICEImputer, EMImputer, MissForestImputer
Matrix completion	imputation_methods.matrix	SoftImputeImputer, PPCAImputer
Neural networks	imputation_methods.neural	AutoencoderImputer, GAINImputer
Ensembles	imputation_methods.ensemble	HybridImputer, StackingImputer, BaggingImputer
Functional API	imputation_methods.functional	mean_impute, median_impute, … (42 functions)
Metrics	imputation_methods.metrics	rmse, mae