Causality Module

The causality module provides tools for comparing models using causal metrics and tests.

Classes

CausalModels

class CausalModels(first_model, second_model)

Bases: object

CausalModels class implementation.

The class is mainly deisgned to implement and include Classification and Regression models of all kinds. The aim is to measure the imapact that one model has on the predictions compared to the second one .

In other words,A/B Testing using causality to measure impact.

Parameters:

• first_model – the first trained model

• second_model – the second trained model

individual_causal_effects

an array of the individual causal effects for the predictions.

Type:

np.ndarray

individual_causal_effects_resid

an array of the individual causal effects for the residuals.

Type:

np.ndarray

average_causal_effect

the effect of the feature on the predictions or outcomes.

Type:

float

average_causal_effect_resid

the average causal effect of the feature on the residuals.

Type:

float

Attributes:

• individual_causal_effects (np.ndarray): an array of the individual causal effects for the predictions

• individual_causal_effects_resid (np.ndarray): an array of the individual causal effects for the residuals

• average_causal_effect (float): the effect of the feature on the predictions or outcomes

• average_causal_effect_resid (float): the average causal effect of the feature on the residuals

individual_causal_effects(X)

The individual causal effects of the predicted outcomes.

Parameters:

X (np.array,pd.DataFrame) – the input values that will be used for prediction.

Returns:

an array of individual causal effects for the predicted values.

Return type:

np.ndarray

individual_causal_effects_resid(X)

The individual causal effects of the residuals. :param X: the input values that will be used for prediction. :type X: np.array,pd.DataFrame

Returns:

an array of individual causal effects for the predicted probabilites.

Return type:

np.ndarray

average_causal_effect(X)

The average difference in the predictions for the two models.

Parameters:

X (np.array,pd.DataFrame) – the input values that will be used for prediction.

Returns:

the average causal effect.

Return type:

float

average_causal_effect_resid(X)

The average difference in the effects of the residuals.

Parameters:

X (np.array,pd.DataFrame) – the input values that will be used for prediction.

Returns:

the average causal effect for the residuals.

Return type:

float

CausalModelsClassifier

class CausalModelsClassifier(first_model, second_model)

Bases: CausalModels

CausalModelClassifier class implementation. The class inherits the CausalModels class

The class is mainly deisgned to implement and include Classification of all kinds. The aim is to measure the imapact that one model has on the predictions compared to the second one .

In other words,A/B Testing using causality to measure impact.

Parameters:

• first_model – the first trained model

• second_model – the second trained model

individual_causal_effects

an array of the individual causal effects for the predictions

Type:

np.ndarray

individual_causal_effects_resid

an array of the individual causal effects for the residuals

Type:

np.ndarray

individual_causal_effects_probs

an array of the individual causal effects for the probabilities

Type:

np.ndarray

average_causal_effect

the average causal effect of the first model

Type:

float

average_causal_effect_resid

the average causal effect of the first model on the residuals

Type:

float

average_causal_effect_probs

the average causal effect of the first model on the predicted probabilities

Type:

float

Attributes:

• individual_causal_effects (np.ndarray): an array of the individual causal effects for the predictions

• individual_causal_effects_resid (np.ndarray): an array of the individual causal effects for the residuals

• individual_causal_effects_probs (np.ndarray): an array of the individual causal effects for the probabilities

• average_causal_effect (float): the average causal effect of the first model

• average_causal_effect_resid (float): the average causal effect of the first model on the residuals

• average_causal_effect_probs (float): the average causal effect of the first model on the predicted probabilities

individual_causal_effects_probs(X)

The individual causal effects of the predicted probabilities

Parameters:

X (np.array,pd.DataFrame) – the input values that will be used for prediction

Returns:

an array of individual causal effects for the predicted probabilities

Return type:

np.ndarray

average_causal_effect_probs(X)

The average difference in the predicted probabilities for the two models

Parameters:

X (np.array,pd.DataFrame) – the input values that will be used for prediction

Returns:

the average causal effect for the predicted probabilities

Return type:

np.ndarray

CausalModelsRegression

class CausalModelsRegression(first_model, second_model)

Bases: CausalModels

CausalModelsRegression class implementation.

The class is mainly deisgned to cover Regression models for tabular data. The aim is to measure the imapact that one model has on the predictions compared to the second one .

In other words,A/B Testing using causality to measure impact

Parameters:

• first_model – the first trained model

• second_model – the second trained model

individual_causal_effects

an array of the individual causal effects for the predictions

Type:

np.ndarray

individual_causal_effects_resid

an array of the individual causal effects for the residuals

Type:

np.ndarray

individual_causal_effects_probs

an array of the individual causal effects for the probabilities

Type:

np.ndarray

average_causal_effect

the average causal effect of the first model

Type:

float

average_causal_effect_resid

the average causal effect of the first model on the residuals

Type:

float

average_causal_effect_probs

the average causal effect of the first model on the predicted probabilities

Type:

float

Attributes:

• individual_causal_effects (np.ndarray): an array of the individual causal effects for the predictions

• individual_causal_effects_resid (np.ndarray): an array of the individual causal effects for the residuals

• individual_causal_effects_probs (np.ndarray): an array of the individual causal effects for the probabilities

• average_causal_effect (float): the average causal effect of the first model

• average_causal_effect_resid (float): the average causal effect of the first model on the residuals

• average_causal_effect_probs (float): the average causal effect of the first model on the predicted probabilities