from __future__ import annotations
from typing import TYPE_CHECKING
import numpy as np
import pandas as pd
import scipy.stats.mstats
from ehrapy._compat import function_2D_only, use_ehrdata
if TYPE_CHECKING:
from collections.abc import Collection
from anndata import AnnData
from ehrdata import EHRData
[docs]
@use_ehrdata(deprecated_after="1.0.0")
@function_2D_only()
def winsorize(
edata: EHRData | AnnData,
vars: Collection[str] = None,
obs_cols: Collection[str] = None,
*,
limits: tuple[float, float] = (0.01, 0.99),
layer: str | None = None,
copy: bool = False,
**kwargs,
) -> EHRData | AnnData | None:
"""Returns a Winsorized version of the input array.
The implementation is based on https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.mstats.winsorize.html
Args:
edata: Central data object.
vars: The features to winsorize.
obs_cols: Columns in obs with features to winsorize.
limits: Tuple of the percentages to cut on each side of the array as floats between 0. and 1.
layer: The layer to operate on.
copy: Whether to return a copy.
**kwargs: Keywords arguments get passed to scipy.stats.mstats.winsorize.
Returns:
Winsorized data object if copy is True.
Examples:
>>> import ehrdata as ed
>>> import ehrapy as ep
>>> edata = ed.dt.mimic_2()
>>> ep.pp.winsorize(edata, vars=["bmi"])
"""
if copy: # pragma: no cover
edata = edata.copy()
obs_cols_set, vars_set = _validate_outlier_input(edata, obs_cols, vars)
if vars_set:
for var in vars_set:
edata_view = edata[:, var]
X = edata_view.X if layer is None else edata_view.layers[layer]
data_array = np.array(X, dtype=float)
winsorized_data = scipy.stats.mstats.winsorize(data_array, limits=limits, nan_policy="omit", **kwargs)
if layer is None:
edata[:, var].X = winsorized_data
else:
edata[:, var].layers[layer] = winsorized_data
if obs_cols_set:
for col in obs_cols_set:
obs_array = edata.obs[col].to_numpy(dtype=float)
winsorized_obs = scipy.stats.mstats.winsorize(obs_array, limits=limits, nan_policy="omit", **kwargs)
edata.obs[col] = pd.Series(winsorized_obs).values
return edata if copy else None
[docs]
@use_ehrdata(deprecated_after="1.0.0")
@function_2D_only()
def clip_quantile(
edata: EHRData | AnnData,
limits: tuple[float, float],
vars: Collection[str] = None,
obs_cols: Collection[str] = None,
*,
layer: str | None = None,
copy: bool = False,
) -> EHRData | AnnData | None:
"""Clips (limits) features.
Given an interval, values outside the interval are clipped to the interval edges.
Args:
edata: Central data object.
limits: Values outside the interval are clipped to the interval edges.
vars: Columns in var with features to clip.
obs_cols: Columns in obs with features to clip
layer: The layer to operate on.
copy: Whether to return a copy of data or not
Returns:
A copy of original data object with clipped features.
Examples:
>>> import ehrdata as ed
>>> import ehrapy as ep
>>> edata = ed.dt.mimic_2()
>>> ep.pp.clip_quantile(edata, limits=(0, 75), vars=["bmi"])
"""
obs_cols, vars = _validate_outlier_input(edata, obs_cols, vars) # type: ignore
if vars:
for var in vars:
edata_view = edata[:, var]
X = edata_view.X if layer is None else edata_view.layers[layer]
X = np.clip(X, limits[0], limits[1])
if layer is None:
edata[:, var].X = X
else:
edata[:, var].layers[layer] = X
if obs_cols:
for col in obs_cols:
obs_array = edata.obs[col].to_numpy()
clipped_array = np.clip(obs_array, limits[0], limits[1])
edata.obs[col] = pd.Series(clipped_array).values
if copy: # pragma: no cover
edata = edata.copy()
return edata if copy else None
def _validate_outlier_input(edata, obs_cols: Collection[str], vars: Collection[str]) -> tuple[set[str], set[str]]:
"""Validates the obs/var columns for outlier preprocessing."""
vars = set(vars) if vars else set()
obs_cols = set(obs_cols) if obs_cols else set()
if vars is not None:
diff = vars - set(edata.var_names)
if len(diff) != 0:
raise ValueError(f"Columns {','.join(var for var in diff)} are not in var_names.")
if obs_cols is not None:
diff = obs_cols - set(edata.obs.columns.values)
if len(diff) != 0:
raise ValueError(f"Columns {','.join(var for var in diff)} are not in obs.")
return obs_cols, vars
