neurostatx.statistics.models

Classes

Name	Description
GAD7Labeler	Assign GAD-7 severity labels from the seven questionnaire items.
PHQ9Labeler	Assign PHQ-9 severity labels from the nine questionnaire items.
Penalty	Regularization penalties for logistic regression.
ScoringMethod	sklearn scoring names accepted by permutation testing.
Solver	sklearn LogisticRegression solver names.

Functions

Name	Description
permutation_testing	Permutation-test an estimator's score and coefficients.
plsr_cv	Fit a PLSR model and choose the component count by cross-validation.

GAD7Labelerclass#

View source

class GAD7Labeler()

Assign GAD-7 severity labels from the seven questionnaire items.

fitmethod#

View source

def fit(X, y=None)

Label GAD-7 item scores.

Delegates to transform.

Parameters

Name	Type	Default	Description
X	pd.DataFrame	required	Item scores of shape (n_samples, 7).
y	pd.Series	None	Unused target, kept for sklearn compatibility.

Returns

labelspd.Series

Severity labels returned by transform.

Examples

>>> import pandas as pd
>>> from neurostatx.statistics.models import GAD7Labeler
>>> X = pd.DataFrame([[0] * 7])
>>> GAD7Labeler().fit(X).iloc[0]
'Subthreshold'

transformmethod#

View source

def transform(X)

Assign a GAD-7 severity label to each row.

Labels are Subthreshold, Mild, Moderate, or Severe.

Parameters

Name	Type	Description
X	pd.DataFrame	Item scores of shape (n_samples, 7).

Returns

labelspd.Series

One severity label per sample.

Examples

>>> import pandas as pd
>>> from neurostatx.statistics.models import GAD7Labeler
>>> X = pd.DataFrame([[0] * 7])
>>> GAD7Labeler().transform(X).iloc[0]
'Subthreshold'

PHQ9Labelerclass#

View source

class PHQ9Labeler()

Assign PHQ-9 severity labels from the nine questionnaire items.

fitmethod#

View source

def fit(X, y=None)

Label PHQ-9 item scores.

Delegates to transform.

Parameters

Name	Type	Default	Description
X	pd.DataFrame	required	Item scores of shape (n_samples, 9).
y	pd.Series	None	Unused target, kept for sklearn compatibility.

Returns

labelspd.Series

Severity labels returned by transform.

Examples

>>> import pandas as pd
>>> from neurostatx.statistics.models import PHQ9Labeler
>>> X = pd.DataFrame([[0] * 9])
>>> PHQ9Labeler().fit(X).iloc[0]
'Subthreshold'

transformmethod#

View source

def transform(X)

Assign a PHQ-9 severity label to each row.

Labels are Subthreshold, Mild, Moderate, Mod-Severe, or Severe.

Parameters

Name	Type	Description
X	pd.DataFrame	Item scores of shape (n_samples, 9).

Returns

labelspd.Series

One severity label per sample.

Examples

>>> import pandas as pd
>>> from neurostatx.statistics.models import PHQ9Labeler
>>> X = pd.DataFrame([[0] * 9])
>>> PHQ9Labeler().transform(X).iloc[0]
'Subthreshold'

Penaltyclass#

View source

class Penalty(StrEnum, Enum)

Bases: StrEnum, Enum

Regularization penalties for logistic regression.

elasticnetattributeclass attributeinstance attribute#

View source

elasticnet = 'elasticnet'

l1attributeclass attributeinstance attribute#

View source

l1 = 'l1'

l2attributeclass attributeinstance attribute#

View source

l2 = 'l2'

ScoringMethodclass#

View source

class ScoringMethod(StrEnum, Enum)

Bases: StrEnum, Enum

sklearn scoring names accepted by permutation testing.

accuracyattributeclass attributeinstance attribute#

View source

accuracy = 'accuracy'

adjusted_mutual_info_scoreattributeclass attributeinstance attribute#

View source

adjusted_mutual_info_score = 'adjusted_mutual_info_score'

adjusted_rand_scoreattributeclass attributeinstance attribute#

View source

adjusted_rand_score = 'adjusted_rand_score'

average_precisionattributeclass attributeinstance attribute#

View source

average_precision = 'average_precision'

balanced_accuracyattributeclass attributeinstance attribute#

View source

balanced_accuracy = 'balanced_accuracy'

completeness_scoreattributeclass attributeinstance attribute#

View source

completeness_score = 'completeness_score'

d2_absolute_error_scoreattributeclass attributeinstance attribute#

View source

d2_absolute_error_score = 'd2_absolute_error_score'

d2_pinball_scoreattributeclass attributeinstance attribute#

View source

d2_pinball_score = 'd2_pinball_score'

d2_tweedie_scoreattributeclass attributeinstance attribute#

View source

d2_tweedie_score = 'd2_tweedie_score'

explained_varianceattributeclass attributeinstance attribute#

View source

explained_variance = 'explained_variance'

f1attributeclass attributeinstance attribute#

View source

f1 = 'f1'

f1_macroattributeclass attributeinstance attribute#

View source

f1_macro = 'f1_macro'

f1_microattributeclass attributeinstance attribute#

View source

f1_micro = 'f1_micro'

f1_samplesattributeclass attributeinstance attribute#

View source

f1_samples = 'f1_samples'

f1_weightedattributeclass attributeinstance attribute#

View source

f1_weighted = 'f1_weighted'

fowlkes_mallows_scoreattributeclass attributeinstance attribute#

View source

fowlkes_mallows_score = 'fowlkes_mallows_score'

homogeneity_scoreattributeclass attributeinstance attribute#

View source

homogeneity_score = 'homogeneity_score'

jaccardattributeclass attributeinstance attribute#

View source

jaccard = 'jaccard'

max_errorattributeclass attributeinstance attribute#

View source

max_error = 'max_error'

mutual_info_scoreattributeclass attributeinstance attribute#

View source

mutual_info_score = 'mutual_info_score'

neg_brier_scoreattributeclass attributeinstance attribute#

View source

neg_brier_score = 'neg_brier_score'

neg_log_lossattributeclass attributeinstance attribute#

View source

neg_log_loss = 'neg_log_loss'

neg_mean_absolute_errorattributeclass attributeinstance attribute#

View source

neg_mean_absolute_error = 'neg_mean_absolute_error'

neg_mean_absolute_percentage_errorattributeclass attributeinstance attribute#

View source

neg_mean_absolute_percentage_error = 'neg_mean_absolute_percentage_error'

neg_mean_gamma_devianceattributeclass attributeinstance attribute#

View source

neg_mean_gamma_deviance = 'neg_mean_gamma_deviance'

neg_mean_poisson_devianceattributeclass attributeinstance attribute#

View source

neg_mean_poisson_deviance = 'neg_mean_poisson_deviance'

neg_mean_squared_errorattributeclass attributeinstance attribute#

View source

neg_mean_squared_error = 'neg_mean_squared_error'

neg_mean_squared_log_errorattributeclass attributeinstance attribute#

View source

neg_mean_squared_log_error = 'neg_mean_squared_log_error'

neg_median_absolute_errorattributeclass attributeinstance attribute#

View source

neg_median_absolute_error = 'neg_median_absolute_error'

neg_root_mean_squared_errorattributeclass attributeinstance attribute#

View source

neg_root_mean_squared_error = 'neg_root_mean_squared_error'

normalized_mutual_info_scoreattributeclass attributeinstance attribute#

View source

normalized_mutual_info_score = 'normalized_mutual_info_score'

precisionattributeclass attributeinstance attribute#

View source

precision = 'precision'

r2attributeclass attributeinstance attribute#

View source

r2 = 'r2'

rand_scoreattributeclass attributeinstance attribute#

View source

rand_score = 'rand_score'

recallattributeclass attributeinstance attribute#

View source

recall = 'recall'

roc_aucattributeclass attributeinstance attribute#

View source

roc_auc = 'roc_auc'

roc_auc_ovoattributeclass attributeinstance attribute#

View source

roc_auc_ovo = 'roc_auc_ovo'

roc_auc_ovo_weightedattributeclass attributeinstance attribute#

View source

roc_auc_ovo_weighted = 'roc_auc_ovo_weighted'

roc_auc_ovrattributeclass attributeinstance attribute#

View source

roc_auc_ovr = 'roc_auc_ovr'

roc_auc_ovr_weightedattributeclass attributeinstance attribute#

View source

roc_auc_ovr_weighted = 'roc_auc_ovr_weighted'

top_k_accuracyattributeclass attributeinstance attribute#

View source

top_k_accuracy = 'top_k_accuracy'

v_measure_scoreattributeclass attributeinstance attribute#

View source

v_measure_score = 'v_measure_score'

Solverclass#

View source

class Solver(StrEnum, Enum)

Bases: StrEnum, Enum

sklearn LogisticRegression solver names.

lbfgsattributeclass attributeinstance attribute#

View source

lbfgs = 'lbfgs'

liblinearattributeclass attributeinstance attribute#

View source

liblinear = 'liblinear'

newton_cgattributeclass attributeinstance attribute#

View source

newton_cg = 'newton-cg'

newton_choleskyattributeclass attributeinstance attribute#

View source

newton_cholesky = 'newton-cholesky'

sagattributeclass attributeinstance attribute#

View source

sag = 'sag'

sagaattributeclass attributeinstance attribute#

View source

saga = 'saga'

permutation_testingfunction#

View source

def permutation_testing(estimator, X, Y, binary=False, nb_permutations=1000, scoring='r2', splits=10, processes=1, verbose=False)

Permutation-test an estimator’s score and coefficients.

Parameters

Name	Type	Default	Description
estimator	estimator	required	Unfitted sklearn-compatible estimator.
X	pd.DataFrame	required	Predictor variables.
Y	pd.DataFrame	required	Dependent variables.
binary	bool	False	Unused flag kept for API compatibility. Defaults to False.
nb_permutations	int	1000	Number of permutations. Defaults to 1000.
scoring	str	'r2'	Scoring name passed to sklearn. Defaults to "r2".
splits	int	10	Number of cross-validation folds. Defaults to 10.
processes	int	1	Number of parallel jobs. Defaults to 1.
verbose	bool	False	If True, increase sklearn Parallel verbosity. Defaults to False.

Returns

modestimator

Fitted estimator from the unpermuted data.

scorefloat

Observed score.

coefarray

Observed coefficients.

perm_scorearray

Permuted scores.

score_pvaluefloat

P-value for the observed score.

perm_coefarray

Permuted coefficients.

coef_pvaluearray

P-values for the coefficients.

Examples

>>> import numpy as np
>>> from sklearn.linear_model import LinearRegression
>>> from neurostatx.statistics.models import permutation_testing
>>> rng = np.random.RandomState(0)
>>> X = rng.rand(20, 2)
>>> Y = X[:, 0] + 0.1 * rng.rand(20)
>>> mod, score, coef, perm_score, score_pvalue, perm_coef, coef_pvalue = (
...     permutation_testing(LinearRegression(), X, Y, nb_permutations=5,
...                         splits=3, processes=1)
... )

plsr_cvfunction#

View source

def plsr_cv(X, Y, nb_comp, max_iter=1000, splits=10, processes=1, verbose=False)

Fit a PLSR model and choose the component count by cross-validation.

Parameters

Name	Type	Default	Description
X	pd.DataFrame	required	Predictor variables.
Y	pd.DataFrame	required	Dependent variables.
nb_comp	int	required	Maximum number of components to evaluate.
max_iter	int	1000	Maximum iterations for PLSR. Defaults to 1000.
splits	int	10	Number of cross-validation folds. Defaults to 10.
processes	int	1	Number of parallel jobs. Defaults to 1.
verbose	bool	False	If True, show a progress bar. Defaults to False.

Returns

plsrPLSRegression

Fitted model using the component count with lowest CV MSE.

mselist

Cross-validated MSE for each component count.

score_cfloat

In-sample R2 score.

score_cvfloat

Cross-validated R2 score.

rscorefloat

Square root of the in-sample R2 score.

mse_cfloat

In-sample mean squared error.

mse_cvfloat

Cross-validated mean squared error.

Examples

>>> import pandas as pd
>>> from neurostatx.statistics.models import plsr_cv
>>> X = pd.DataFrame({"a": [1, 2, 3, 4], "b": [2, 1, 4, 3]})
>>> Y = pd.DataFrame({"y": [1.1, 1.9, 3.2, 3.8]})
>>> plsr, mse, score_c, score_cv, rscore, mse_c, mse_cv = plsr_cv(
...     X, Y, nb_comp=2, splits=2
... )