neurostatx.statistics.utils

Functions

Name	Description
KNNimputation	Impute missing values in df from a complete reference table.
apply_various_models	Apply a fitted factor or sklearn transformer to df.

KNNimputationfunction#

View source

def KNNimputation(ref_df, df, n_neighbors=5, weights='distance', metric='nan_euclidean', keep_all_features=True)

Impute missing values in df from a complete reference table.

Fits sklearn KNNImputer on ref_df and applies it to df. Both tables must share the same columns, and ref_df should contain no missing values.

Parameters

Name	Type	Default	Description
ref_df	pd.DataFrame	required	Complete reference dataset used to learn neighbor relationships.
df	pd.DataFrame	required	Dataset to impute.
n_neighbors	int	5	Number of neighbors. Defaults to 5.
weights	str	'distance'	Neighbor weighting, "uniform" or "distance". Defaults to "distance".
metric	str	'nan_euclidean'	Distance metric. Defaults to "nan_euclidean".
keep_all_features	bool	True	If True, columns that are entirely missing are still imputed. Defaults to True.

Returns

outpd.DataFrame

Imputed dataset.

Examples

>>> import pandas as pd
>>> import numpy as np
>>> from neurostatx.statistics.utils import KNNimputation
>>> ref = pd.DataFrame({"a": [1.0, 2.0, 3.0], "b": [2.0, 4.0, 6.0]})
>>> df = pd.DataFrame({"a": [1.5, np.nan], "b": [3.0, 5.0]})
>>> KNNimputation(ref, df, n_neighbors=2).isna().sum().sum()
0

apply_various_modelsfunction#

View source

def apply_various_models(df, mod)

Apply a fitted factor or sklearn transformer to df.

Supports semopy models, sklearn transformers, and factor_analyzer models.

Parameters

Name	Type	Description
df	pd.DataFrame	Data to transform.
mod	object	Fitted model exposing predict_factors or transform.

Returns

ypd.DataFrame

Transformed factor or component scores.

Examples

>>> import pandas as pd
>>> from sklearn.decomposition import PCA
>>> from neurostatx.statistics.utils import apply_various_models
>>> df = pd.DataFrame({"a": [1.0, 2.0, 3.0], "b": [2.0, 4.0, 6.0]})
>>> pca = PCA(n_components=1).fit(df)
>>> apply_various_models(df, pca).shape
(3, 1)