EconML

https://www.pywhy.org/EconML
Admin
EconML is a Python package that applies the power of machine learning techniques to estimate...
Tokens:1,458,289
Snippets:7,567
Trust Score:8.7
Update:1 day ago
Show doc for...
Tokens:
Raw
### RegressionForest example: setup and training

Source: https://www.pywhy.org/EconML/_autosummary/econml.grf.RegressionForest.html

Imports necessary libraries, generates synthetic regression data, splits into train/test sets, and initializes a RegressionForest with specified parameters.

```python
import numpy as np
from econml.grf import RegressionForest
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split

np.set_printoptions(suppress=True)
np.random.seed(123)
X, y = make_regression(n_samples=1000, n_features=4, n_informative=2,
                       random_state=0, shuffle=False)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.5)
regr = RegressionForest(max_depth=None, random_state=0,
                        n_estimators=1000)
```

--------------------------------

### Example: Implementing a Simple RLearner

Source: https://www.pywhy.org/EconML/_autosummary/econml.dml._rlearner.html

Example code demonstrating how to subclass _RLearner to implement a simple double machine learning method with custom model selectors and final model.

```APIDOC
## Example: Implementing a Simple RLearner

### Description
This example implements a very simple version of the double machine learning method on top of the _RLearner class, for expository purposes. For a more elaborate implementation, see DML and its child classes.

### Code Example
```python
import numpy as np
from sklearn.linear_model import LinearRegression
from econml.dml._rlearner import _RLearner
from econml.sklearn_extensions.model_selection import SingleModelSelector
from sklearn.base import clone

class ModelFirst:
    def __init__(self, model):
        self._model = clone(model, safe=False)
    def fit(self, X, W, Y, sample_weight=None):
        self._model.fit(np.hstack([X, W]), Y)
        return self
    def predict(self, X, W):
        return self._model.predict(np.hstack([X, W]))

class ModelSelector(SingleModelSelector):
    def __init__(self, model):
        self._model = ModelFirst(model)
    def train(self, is_selecting, folds, X, W, Y, sample_weight=None):
        self._model.fit(X, W, Y, sample_weight=sample_weight)
        return self
    @property
    def best_model(self):
        return self._model
    @property
    def best_score(self):
        return 0

class ModelFinal:
    def fit(self, X, T, T_res, Y_res, sample_weight=None, freq_weight=None, sample_var=None):
        self.model = LinearRegression(fit_intercept=False).fit(X * T_res.reshape(-1, 1), Y_res)
        return self
    def predict(self, X):
        return self.model.predict(X)

class RLearner(_RLearner):
    def _gen_model_y(self):
        return ModelSelector(LinearRegression())
    def _gen_model_t(self):
        return ModelSelector(LinearRegression())
    def _gen_rlearner_model_final(self):
        return ModelFinal()

np.random.seed(123)
X = np.random.normal(size=(1000, 3))
y = X[:, 0] + X[:, 1] + np.random.normal(0, 0.01, size=(1000,))
est = RLearner(cv=2, discrete_outcome=False, discrete_treatment=False,
               treatment_featurizer=None, categories='auto', random_state=None)
est.fit(y, X[:, 0], X=np.ones((X.shape[0], 1)), W=X[:, 1:])
```
```

--------------------------------

### DMLOrthoForest Usage Example

Source: https://www.pywhy.org/EconML/spec/estimation/forest.html

This example demonstrates how to instantiate and fit a DMLOrthoForest estimator and then use its effect method to estimate treatment effects on new data points.

```APIDOC
## DMLOrthoForest

### Description
DMLOrthoForest is an orthogonal random forest estimator for heterogeneous treatment effects. It uses double machine learning with random forests to estimate conditional average treatment effects (CATE).

### Usage

#### Import
```python
from econml.orf import DMLOrthoForest
```

#### Constructor Parameters
- `n_trees` (int) - Number of trees in the forest.
- `max_depth` (int) - Maximum depth of each tree.
- `subsample_ratio` (float) - Ratio of samples to use for each tree.
- `model_T` (estimator) - Model for the treatment regression.
- `model_Y` (estimator) - Model for the outcome regression.

#### Methods

##### fit(Y, T, X=None, W=None)
Fits the estimator to the data.

**Parameters:**
- `Y` (array-like) - Outcome vector.
- `T` (array-like) - Treatment vector.
- `X` (array-like, optional) - Features for heterogeneity.
- `W` (array-like, optional) - Confounders.

**Returns:**
- The fitted estimator instance.

##### effect(X=None)
Estimates the conditional average treatment effect (CATE) at the given feature points.

**Parameters:**
- `X` (array-like, optional) - Feature points at which to evaluate the effect. If None, uses the training features.

**Returns:**
- `ndarray` - Estimated treatment effects.

### Example
```python
import numpy as np
import sklearn
from econml.orf import DMLOrthoForest

np.random.seed(123)
T = np.array([0, 1]*60)
W = np.array([0, 1, 1, 0]*30).reshape(-1, 1)
Y = (.2 * W[:, 0] + 1) * T + .5
est = DMLOrthoForest(n_trees=1, max_depth=1, subsample_ratio=1,
                     model_T=sklearn.linear_model.LinearRegression(),
                     model_Y=sklearn.linear_model.LinearRegression())
est.fit(Y, T, X=W, W=W)
print(est.effect(W[:2]))
```
```

--------------------------------

### DomainAdaptationLearner usage example

Source: https://www.pywhy.org/EconML/_modules/econml/metalearners/_metalearners.html

Example of using DomainAdaptationLearner with LinearRegression models. It generates synthetic data, fits the estimator, and prints the effect for the first three samples. The output shows the estimated effects.

```python
import numpy as np
import scipy.special
np.set_printoptions(suppress=True)
```

```python
from econml.metalearners import DomainAdaptationLearner
from sklearn.linear_model import LinearRegression

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = DomainAdaptationLearner(
    models=LinearRegression(),
    final_models=LinearRegression()
)
est.fit(y, T, X=X)
```

```python
>>> est.effect(X[:3])
array([0.51237..., 1.99866..., 0.68552...])
```

--------------------------------

### Simple example with default models

Source: https://www.pywhy.org/EconML/_autosummary/econml.iv.dr.IntentToTreatDRIV.html

This example demonstrates how to use IntentToTreatDRIV with default models. It defines a data generating process (dgp) and a true heterogeneity function, then fits the estimator on synthetic data.

```python
from econml.iv.dr import IntentToTreatDRIV

# Define the data generation functions
def dgp(n, p, true_fn):
    X = np.random.normal(0, 1, size=(n, p))
    Z = np.random.binomial(1, 0.5, size=(n,))
    nu = np.random.uniform(0, 10, size=(n,))
    coef_Z = 0.8
    C = np.random.binomial(
        1, coef_Z * scipy.special.expit(0.4 * X[:, 0] + nu)
    )  # Compliers when recomended
    C0 = np.random.binomial(
        1, 0.06 * np.ones(X.shape[0])
    )  # Non-compliers when not recommended
    T = C * Z + C0 * (1 - Z)
    y = true_fn(X) * T + 2 * nu + 5 * (X[:, 3] > 0) + 0.1 * np.random.uniform(0, 1, size=(n,))
    return y, T, Z, X

def true_heterogeneity_function(X):
    return 5 * X[:, 0]

np.random.seed(123)
y, T, Z, X = dgp(1000, 5, true_heterogeneity_function)
est = IntentToTreatDRIV()
est.fit(Y=y, T=T, Z=Z, X=X)
```

```python
>>> est.effect(X[:3])
array([-4.52684...,  6.38767..., -2.67082...])
```

--------------------------------

### Import numpy and set print options for DynamicDML example

Source: https://www.pywhy.org/EconML/_modules/econml/panel/dml/_dml.html

Hidden setup code that imports numpy and suppresses scientific notation for the example output.

```python
import numpy as np
np.set_printoptions(suppress=True)
```

--------------------------------

### DomainAdaptationLearner.__init__()

Source: https://www.pywhy.org/EconML/

Initializes the DomainAdaptationLearner.

```APIDOC
## DomainAdaptationLearner.__init__()

### Description
Initializes the DomainAdaptationLearner.

### Method
__init__

### Parameters
No parameters documented.

### Returns
No return value documented.
```

--------------------------------

### Import numpy and scipy.special for CausalForestDML example

Source: https://www.pywhy.org/EconML/_modules/econml/dml/causal_forest.html

Hidden setup code that imports numpy and scipy.special and sets numpy print options to suppress scientific notation. This is required for the subsequent example to run correctly.

```python
import numpy as np
import scipy.special
np.set_printoptions(suppress=True)
```

--------------------------------

### GRFTree.init

Source: https://www.pywhy.org/EconML/_autosummary/econml.grf._base_grftree.GRFTree.html

Initializes this instance.

```APIDOC
## GRFTree.init

### Description
Initializes this instance.

### Method
init

### Signature
`init()`

### Returns
None.
```

--------------------------------

### DMLOrthoForest.__init__()

Source: https://www.pywhy.org/EconML/index.html

Initializes a DMLOrthoForest instance.

```APIDOC
## DMLOrthoForest.__init__()

### Description
Initializes a DMLOrthoForest instance.

### Method
Constructor

### Endpoint
`DMLOrthoForest.__init__()`

### Parameters
No parameters documented.

### Returns
A new DMLOrthoForest instance.
```

--------------------------------

### strata_from_discrete_arrays: Combine discrete arrays for stratification

Source: https://www.pywhy.org/EconML/_modules/econml/utilities.html

Combines multiple discrete arrays into a single array for stratification purposes. Each distinct combination of the input arrays gets its own label. The example shows input arrays and the resulting output array.

```python
[docs]
def strata_from_discrete_arrays(arrs):
    """
    Combine multiple discrete arrays into a single array for stratification purposes.

    For example, if arrs is
    `[[0 1 2 0 1 2 0 1 2 0 1 2],
    [0 1 0 1 0 1 0 1 0 1 0 1],
    [0 0 0 0 0 0 1 1 1 1 1 1]]`
    then output will be
    `[0 8 4 6 2 10 1 9 5 7 3 11]`

    Every distinct combination of these discrete arrays will have it's own label.
    """
    if not arrs:
        return None

    curr_array = np.zeros(shape=np.ravel(arrs[0]).shape, dtype='int')

    for arr in arrs:
        enc = LabelEncoder()
        temp = enc.fit_transform(np.ravel(arr))
        curr_array = temp + curr_array * len(enc.classes_)

    return curr_array
```

--------------------------------

### Example: Output of discrete treatment example

Source: https://www.pywhy.org/EconML/_autosummary/econml._ortho_learner.html

Shows the output of the discrete treatment example, including the score_, const_marginal_effect(), effect(), and the coefficient of the final model.

```python
>>> est.score_
np.float64(0.00672978...)
>>> est.const_marginal_effect()
array([[1.008402...]])
>>> est.effect()
array([1.008402...])
>>> est.score(y, T, W=W)
np.float64(0.00310431...)
>>> est.ortho_learner_model_final_.model.coef_[0]
np.float64(1.00840240...)
```

--------------------------------

### LinearDRLearner Example Usage

Source: https://www.pywhy.org/EconML/_modules/econml/dr/_drlearner.html

Example demonstrating how to use LinearDRLearner to estimate treatment effects and extract coefficients.

```APIDOC
## LinearDRLearner Example Usage

### Description
Example demonstrating how to use LinearDRLearner to estimate treatment effects and extract coefficients.

### Example
```python
import numpy as np
import scipy.special
from econml.dr import LinearDRLearner

np.random.seed(123)
X = np.random.normal(size=(1000, 3))
T = np.random.binomial(2, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = LinearDRLearner()
est.fit(y, T, X=X, W=None)

est.effect(X[:3])
# array([ 0.432365...,  0.359694..., -0.085428...])

est.effect_interval(X[:3])
# (array([ 0.084048..., -0.177951... , -0.734747...]),
#  array([0.780683..., 0.897341..., 0.563889...]))

est.coef_(T=1)
# array([ 0.450666..., -0.008821...,  0.075271...])

est.coef__interval(T=1)
# (array([ 0.156245..., -0.252216..., -0.159709...]),
#  array([0.745086..., 0.234572..., 0.310252...]))

est.intercept_(T=1)
# np.float64(0.909121...)

est.intercept__interval(T=1)
# (np.float64(0.668518...), np.float64(1.149723...))
```

### Attributes
- **score_** (float) - The MSE in the final doubly robust potential outcome regressions, i.e., \frac{1}{n_t} \sum_{t=1}^{n_t} \frac{1}{n} \sum_{i=1}^n (Y_{i, t}^{DR} - \hat{\theta}_t(X_i))^2, where n_t is the number of treatments (excluding control). If sample_weight is not None at fit time, then a weighted average across samples is returned.
```

--------------------------------

### XLearner.__init__

Source: https://www.pywhy.org/EconML/_autosummary/econml.metalearners.XLearner.html

Initializes the XLearner estimator with outcome models, CATE models, propensity model, and treatment categories.

```APIDOC
## XLearner.__init__

### Description
Initializes the XLearner estimator with the specified outcome models, CATE models, propensity model, and treatment categories.

### Method
Constructor

### Signature
`__init__(*, models, cate_models=None, propensity_model=LogisticRegression(), categories='auto', allow_missing=False)`

### Parameters
- **models** (outcome estimators for both control units and treatment units) - Required - It can be a single estimator applied to all the control and treatment units or a tuple/list of estimators with one estimator per treatment (including control). Must implement fit and predict methods.
- **cate_models** (estimator for pseudo-treatment effects on control and treatments) - Optional - It can be a single estimator applied to all the control and treatments or a tuple/list of estimators with one estimator per treatment (including control). If None, it will be same models as the outcome estimators. Must implement fit and predict methods.
- **propensity_model** (estimator for the propensity function) - Optional - Must implement fit and predict_proba methods. The fit method must be able to accept X and T, where T is a shape (n, ) array.
- **categories** ('auto' or list, default 'auto') - Optional - The categories to use when encoding discrete treatments (or 'auto' to use the unique sorted values). The first category will be treated as the control treatment.
- **allow_missing** (bool) - Optional - Whether to allow missing values in X. If True, will need to supply models, cate_models, and propensity_model that can handle missing values.

### Example
```python
from econml.metalearners import XLearner
from sklearn.linear_model import LinearRegression

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = XLearner(models=LinearRegression())
est.fit(y, T, X=X)
```
```

--------------------------------

### Simple DMLOrthoForest example

Source: https://www.pywhy.org/EconML/spec/estimation/forest.html

Creates synthetic data and fits a DMLOrthoForest with linear models for treatment and outcome. Prints the estimated effects for the first two samples.

```python
>>> T = np.array([0, 1]*60)
>>> W = np.array([0, 1, 1, 0]*30).reshape(-1, 1)
>>> Y = (.2 * W[:, 0] + 1) * T + .5
>>> est = DMLOrthoForest(n_trees=1, max_depth=1, subsample_ratio=1,
...                      model_T=sklearn.linear_model.LinearRegression(),
...                      model_Y=sklearn.linear_model.LinearRegression())
>>> est.fit(Y, T, X=W, W=W)
<econml.orf._ortho_forest.DMLOrthoForest object at 0x...>
>>> print(est.effect(W[:2]))
[1.00...  1.19...]
```

--------------------------------

### CausalForestDML Usage Example

Source: https://www.pywhy.org/EconML/_autosummary/econml.dml.CausalForestDML.html

Example of fitting a CausalForestDML model and using the effect and effect_interval methods to estimate treatment effects.

```APIDOC
## CausalForestDML Usage Example

### Description
Demonstrates how to fit a CausalForestDML model and use its effect and effect_interval methods.

### Example
```python
from econml.dml import CausalForestDML

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = CausalForestDML(discrete_treatment=True)
est.fit(y, T, X=X, W=None)

>>> est.effect(X[:3])
array([0.62769..., 1.64575..., 0.68497...])
>>> est.effect_interval(X[:3])
(array([0.18798..., 1.17144..., 0.10788...]),
array([1.06739..., 2.12006..., 1.26205...]))
```
```

--------------------------------

### Bootstrap Inference Example

Source: https://www.pywhy.org/EconML/spec/inference.html

Example showing how to enable bootstrap inference in an EconML estimator and compute point estimates and confidence intervals.

```APIDOC
## Bootstrap Inference

### Description
Enable bootstrap-based confidence intervals for any EconML estimator by passing `inference='bootstrap'` or `inference=BootstrapInference(n_bootstrap_samples=100, n_jobs=-1)`. The intervals are calculated by training multiple versions of the estimator on bootstrap subsamples with replacement and then taking quantiles of the estimate distribution.

### Usage
```python
from econml.dml import NonParamDML
from sklearn.ensemble import RandomForestRegressor
est = NonParamDML(model_y=RandomForestRegressor(n_estimators=10, min_samples_leaf=10),
                            model_t=RandomForestRegressor(n_estimators=10, min_samples_leaf=10),
                            model_final=RandomForestRegressor(n_estimators=10, min_samples_leaf=10))
est.fit(y, t, X=X, W=W, inference='bootstrap')
point = est.const_marginal_effect(X)
lb, ub = est.const_marginal_effect_interval(X, alpha=0.05)
```

### Parameters
- `inference` (str or BootstrapInference) - Required - Either the string `'bootstrap'` or an instance of `BootstrapInference` with `n_bootstrap_samples` and `n_jobs` parameters.

### Methods
- `fit(y, t, X=X, W=W, inference='bootstrap')` - Fits the estimator with bootstrap inference enabled.
- `const_marginal_effect(X)` - Returns the point estimate of the constant marginal effect.
- `const_marginal_effect_interval(X, alpha=0.05)` - Returns the lower and upper bounds of the confidence interval at the specified alpha level.

### Notes
- The `BootstrapInference` class is referenced as `econml.inference.BootstrapInference`.
- The `BootstrapEstimator` class (referenced as `econml.inference._bootstrap.BootstrapEstimator`) provides more details on the bootstrap procedure.
```

--------------------------------

### init()

Source: https://www.pywhy.org/EconML/_autosummary/econml.grf._base_grftree.GRFTree.html

Initializes this instance. This method should be called before fit. It is a pre-fit step to avoid parallelism issues and to keep the starting seed for replicating randomness.

```APIDOC
## init()

### Description
Initializes this instance. This method should be called before fit. It is a pre-fit step to avoid parallelism issues and to keep the starting seed for replicating randomness.

### Method
`init()`

```

--------------------------------

### LinearDRLearner Usage Example

Source: https://www.pywhy.org/EconML/spec/estimation/dr.html

Example of fitting a LinearDRLearner and computing the conditional average treatment effect (CATE) for given treatment values.

```APIDOC
## LinearDRLearner

### Description
Fits a linear double machine learning model to estimate conditional average treatment effects (CATE) as a linear function of features X.

### Method
fit

### Signature
`est.fit(y, T, X=X, W=W)`

### Parameters
- **y** (array-like) - Required - Outcome variable.
- **T** (array-like) - Required - Treatment variable.
- **X** (array-like) - Optional - Features for heterogeneity.
- **W** (array-like) - Optional - Controls/confounders.

### Returns
Fitted estimator object.

### Method
effect

### Signature
`est.effect(X, T0=t0, T1=t1)`

### Parameters
- **X** (array-like) - Required - Features for which to compute the effect.
- **T0** (optional) - Baseline treatment.
- **T1** (optional) - Treatment to compare against.

### Returns
Estimated conditional average treatment effect (CATE) for each sample in X.

### Example
```python
from econml.dr import LinearDRLearner
est = LinearDRLearner()
est.fit(y, T, X=X, W=W)
est.effect(X, T0=t0, T1=t1)
```
```

--------------------------------

### PolicyForest.__init__()

Source: https://www.pywhy.org/EconML/index.html

Initializes a PolicyForest instance.

```APIDOC
## PolicyForest.__init__()

### Description
Initializes a PolicyForest instance.

### Method
__init__

### Parameters
No parameters documented.

### Returns
No return value documented.
```

--------------------------------

### TLearner basic usage example

Source: https://www.pywhy.org/EconML/_autosummary/econml.metalearners.TLearner.html

Shows how to instantiate TLearner with a linear regression model, fit it on synthetic data, and compute the treatment effect for the first three samples.

```python
from econml.metalearners import TLearner
from sklearn.linear_model import LinearRegression

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = TLearner(models=LinearRegression())
est.fit(y, T, X=X)
```

```python
>>> est.effect(X[:3])
array([0.58547..., 1.82860..., 0.78379...])
```

--------------------------------

### DROrthoForest Usage Example

Source: https://www.pywhy.org/EconML/spec/estimation/forest.html

This example demonstrates how to instantiate and fit a DROrthoForest estimator and then use its effect method to estimate treatment effects on new data points.

```APIDOC
## DROrthoForest

### Description
DROrthoForest is an orthogonal random forest estimator for heterogeneous treatment effects with binary or continuous treatments, using a propensity model for the treatment.

### Usage

#### Import
```python
from econml.orf import DROrthoForest
```

#### Constructor Parameters
- `n_trees` (int) - Number of trees in the forest.
- `max_depth` (int) - Maximum depth of each tree.
- `subsample_ratio` (float) - Ratio of samples to use for each tree.
- `propensity_model` (estimator) - Model for the propensity score.
- `model_Y` (estimator) - Model for the outcome regression.

#### Methods

##### fit(Y, T, X=None, W=None)
Fits the estimator to the data.

**Parameters:**
- `Y` (array-like) - Outcome vector.
- `T` (array-like) - Treatment vector.
- `X` (array-like, optional) - Features for heterogeneity.
- `W` (array-like, optional) - Confounders.

**Returns:**
- The fitted estimator instance.

##### effect(X=None)
Estimates the conditional average treatment effect (CATE) at the given feature points.

**Parameters:**
- `X` (array-like, optional) - Feature points at which to evaluate the effect. If None, uses the training features.

**Returns:**
- `ndarray` - Estimated treatment effects.

### Example
```python
import numpy as np
import sklearn
from econml.orf import DROrthoForest

np.random.seed(123)
T = np.array([0, 1]*60)
W = np.array([0, 1, 1, 0]*30).reshape(-1, 1)
Y = (.2 * W[:, 0] + 1) * T + .5
est = DROrthoForest(n_trees=1, max_depth=1, subsample_ratio=1,
                    propensity_model=sklearn.linear_model.LogisticRegression(),
                    model_Y=sklearn.linear_model.LinearRegression())
est.fit(Y, T, X=W, W=W)
print(est.effect(W[:2]))
```
```

--------------------------------

### DRIV example with default models

Source: https://www.pywhy.org/EconML/_autosummary/econml.iv.dr.DRIV.html

Shows how to instantiate and fit a DRIV estimator with discrete treatment and instrument, using a synthetic data generating process. The example defines helper functions for data generation and true heterogeneity, then fits the model and prints effects for the first three samples.

```python
from econml.iv.dr import DRIV

# Define the data generation functions
def dgp(n, p, true_fn):
    X = np.random.normal(0, 1, size=(n, p))
    Z = np.random.binomial(1, 0.5, size=(n,))
    nu = np.random.uniform(0, 10, size=(n,))
    coef_Z = 0.8
    C = np.random.binomial(
        1, coef_Z * scipy.special.expit(0.4 * X[:, 0] + nu)
    )  # Compliers when recomended
    C0 = np.random.binomial(
        1, 0.06 * np.ones(X.shape[0])
    )  # Non-compliers when not recommended
    T = C * Z + C0 * (1 - Z)
    y = true_fn(X) * T + 2 * nu + 5 * (X[:, 3] > 0) + 0.1 * np.random.uniform(0, 1, size=(n,))
    return y, T, Z, X

def true_heterogeneity_function(X):
    return 5 * X[:, 0]

np.random.seed(123)
y, T, Z, X = dgp(1000, 5, true_heterogeneity_function)
est = DRIV(discrete_treatment=True, discrete_instrument=True)
est.fit(Y=y, T=T, Z=Z, X=X)
```

```python
>>> est.effect(X[:3])
array([-4.15076...,  5.99286..., -2.86512...])
```

--------------------------------

### LinearDML Usage Example

Source: https://www.pywhy.org/EconML/spec/estimation/dml.html

Shows how to fit a LinearDML model and estimate the treatment effect. The example uses generic model_y and model_t parameters, which can be any scikit-learn regressor.

```python
est.fit(y, t, X=X, W=W)
point = est.effect(X, T0=t0, T1=t1)
```

--------------------------------

### Example: Using _OrthoLearner for continuous treatments

Source: https://www.pywhy.org/EconML/_autosummary/econml._ortho_learner.html

Shows how to use the _OrthoLearner class for continuous treatments. The example demonstrates fitting the model and accessing various attributes and methods such as score_, const_marginal_effect(), effect(), and the underlying linear regression model.

```python
>>> est.score_
np.float64(0.00756830...)
>>> est.const_marginal_effect()
np.float64(1.02364992...)
>>> est.effect()
array([1.023649...])
>>> est.effect(T0=0, T1=10)
array([10.236499...])
>>> est.score(y, X[:, 0], W=X[:, 1:])
np.float64(0.00727995...)
>>> est.ortho_learner_model_final_.model
LinearRegression(fit_intercept=False)
>>> est.ortho_learner_model_final_.model.coef_
array([1.023649...])
```

--------------------------------

### Simple DRLearner example with default models

Source: https://www.pywhy.org/EconML/_autosummary/econml.dr.DRLearner.html

Shows how to instantiate and fit a DRLearner with default models on synthetic data. Sets a random seed, generates features X, treatment T, and outcome y, then fits the estimator.

```python
from econml.dr import DRLearner

np.random.seed(123)
X = np.random.normal(size=(1000, 3))
T = np.random.binomial(2, scipy.special.expit(X[:, 0]))
sigma = 0.001
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(0, sigma, size=(1000,))
est = DRLearner()
est.fit(y, T, X=X, W=None)
```

--------------------------------

### _OrthoLearner Class Example

Source: https://www.pywhy.org/EconML/_autosummary/econml._ortho_learner.html

Example code demonstrating how to subclass _OrthoLearner to implement a simple double machine learning method, including custom nuisance and final models.

```APIDOC
## _OrthoLearner

### Description
Base class for orthogonal learners. This example shows how to subclass it to implement a simple double machine learning method.

### Parameters
- **categories** ('auto' or list) - The categories to use when encoding discrete treatments (or 'auto' to use the unique sorted values). The first category will be treated as the control treatment.
- **cv** (int, cross-validation generator or an iterable) - Determines the cross-validation splitting strategy. Possible inputs: None (default 3-fold), integer (number of folds), CV splitter, or an iterable yielding (train, test) splits. For integer/None inputs, StratifiedKFold is used for discrete treatments, else KFold with random shuffle.
- **random_state** (int, RandomState instance or None) - Seed for random number generator, or RandomState instance, or None to use np.random.
- **mc_iters** (int, optional) - Number of times to rerun the first stage models to reduce variance of nuisances.
- **mc_agg** ({'mean', 'median'}, default 'mean') - How to aggregate nuisance values across mc_iters iterations.
- **allow_missing** (bool) - Whether to allow missing values in X, W. If True, nuisance models must handle missing values.
- **use_ray** (bool, default False) - Whether to use ray to parallelize the cross-fitting step.
- **ray_remote_func_options** (dict, default None) - Options to pass to the ray.remote decorator.

### Example Code
```python
import numpy as np
from sklearn.linear_model import LinearRegression
from econml._ortho_learner import _OrthoLearner

class ModelNuisance:
    def __init__(self, model_t, model_y):
        self._model_t = model_t
        self._model_y = model_y
    def train(self, is_selecting, folds, Y, T, W=None):
        self._model_t.fit(W, T)
        self._model_y.fit(W, Y)
        return self
    def predict(self, Y, T, W=None):
        return Y - self._model_y.predict(W), T - self._model_t.predict(W)

class ModelFinal:
    def __init__(self):
        return
    def fit(self, Y, T, W=None, nuisances=None):
        Y_res, T_res = nuisances
        self.model = LinearRegression(fit_intercept=False).fit(T_res.reshape(-1, 1), Y_res)
        return self
    def predict(self, X=None):
        return self.model.coef_[0]
    def score(self, Y, T, W=None, nuisances=None):
        Y_res, T_res = nuisances
        return np.mean((Y_res - self.model.predict(T_res.reshape(-1, 1)))**2)

class OrthoLearner(_OrthoLearner):
    def _gen_ortho_learner_model_nuisance(self):
        return ModelNuisance(LinearRegression(), LinearRegression())
    def _gen_ortho_learner_model_final(self):
        return ModelFinal()

np.random.seed(123)
X = np.random.normal(size=(100, 3))
y = X[:, 0] + X[:, 1] + np.random.normal(0, 0.1, size=(100,))
est = OrthoLearner(cv=2, discrete_outcome=False, discrete_treatment=False, treatment_featurizer=None,
                   discrete_instrument=False, categories='auto', random_state=None)
est.fit(y, X[:, 0], W=X[:, 1:])
```
```

--------------------------------

### BootstrapInference.__init__()

Source: https://www.pywhy.org/EconML/

Initializes the BootstrapInference object.

```APIDOC
## BootstrapInference.__init__()

### Description
Initializes the BootstrapInference object.

### Method
__init__()

### Parameters
None documented in source.
```

--------------------------------

### XLearner simple example

Source: https://www.pywhy.org/EconML/_autosummary/econml.metalearners.XLearner.html

A basic example of using XLearner with a LinearRegression model. It generates synthetic data, fits the model, and then predicts the effect for the first three samples.

```python
from econml.metalearners import XLearner
from sklearn.linear_model import LinearRegression

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = XLearner(models=LinearRegression())
est.fit(y, T, X=X)
```

```python
>>> est.effect(X[:3])
array([0.58547..., 1.82860..., 0.78379...])
```

--------------------------------

### SparseLinearDRIV simple example with default models

Source: https://www.pywhy.org/EconML/_autosummary/econml.iv.dr.SparseLinearDRIV.html

This example shows how to use SparseLinearDRIV with default models. It defines a data generating process with a binary instrument and treatment, fits the model, and then evaluates the effect and coefficients. The output shows the estimated effects and confidence intervals.

```python
from econml.iv.dr import SparseLinearDRIV

# Define the data generation functions
def dgp(n, p, true_fn):
    X = np.random.normal(0, 1, size=(n, p))
    Z = np.random.binomial(1, 0.5, size=(n,))
    nu = np.random.uniform(0, 10, size=(n,))
    coef_Z = 0.8
    C = np.random.binomial(
        1, coef_Z * scipy.special.expit(0.4 * X[:, 0] + nu)
    )  # Compliers when recomended
    C0 = np.random.binomial(
        1, 0.06 * np.ones(X.shape[0])
    )  # Non-compliers when not recommended
    T = C * Z + C0 * (1 - Z)
    y = true_fn(X) * T + 2 * nu + 5 * (X[:, 3] > 0) + 0.1 * np.random.uniform(0, 1, size=(n,))
    return y, T, Z, X

def true_heterogeneity_function(X):
    return 5 * X[:, 0]

np.random.seed(123)
y, T, Z, X = dgp(1000, 5, true_heterogeneity_function)
est = SparseLinearDRIV(discrete_treatment=True, discrete_instrument=True)
est.fit(Y=y, T=T, Z=Z, X=X)
```

```python
>>> est.effect(X[:3])
array([-4.23922...,  5.89221..., -3.01203...])
>>> est.effect_interval(X[:3])
(array([-6.99782...,  1.96348..., -5.41958...]),
array([-1.48062...,  9.82093..., -0.60449...]))
>>> est.coef_
array([ 4.65817...,  0.94689... ,  0.18314... ,  0.23012..., -0.40375...])
>>> est.coef__interval()
(array([ 3.51645..., -0.20838..., -0.99568..., -0.89395..., -1.58517...]),
array([5.79989..., 2.10218..., 1.36197..., 1.35419..., 0.77767...]))
>>> est.intercept_
-0.06535...
>>> est.intercept__interval()
(-1.20712..., 1.07641...)
```

--------------------------------

### Import libraries for OrthoForest examples

Source: https://www.pywhy.org/EconML/spec/estimation/forest.html

Imports numpy, sklearn, and the OrthoForest classes. Sets a random seed for reproducibility.

```python
import numpy as np
import sklearn
from econml.orf import DMLOrthoForest, DROrthoForest
np.random.seed(123)
```

--------------------------------

### residuals_

Source: https://www.pywhy.org/EconML/_autosummary/econml.dml.SparseLinearDML.html

Get the residuals.

```APIDOC
## residuals_

### Description
Get the residuals.

### Returns
- **residuals_** - The residuals.
```

--------------------------------

### Simple example with default models using DynamicDML

Source: https://www.pywhy.org/EconML/_autosummary/econml.panel.dml.DynamicDML.html

Shows how to create synthetic panel data and fit a DynamicDML model with default settings. The example sets a random seed, creates panel groups, features X, treatments T, and outcome y, then fits the model with inference='auto'.

```python
from econml.panel.dml import DynamicDML

np.random.seed(123)

n_panels = 100  # number of panels
n_periods = 3  # number of time periods per panel
n = n_panels * n_periods
groups = np.repeat(a=np.arange(n_panels), repeats=n_periods, axis=0)
X = np.random.normal(size=(n, 1))
T = np.random.normal(size=(n, 2))
y = np.random.normal(size=(n, ))
est = DynamicDML()
est.fit(y, T, X=X, W=None, groups=groups, inference="auto")
```

--------------------------------

### IVForest.__init__

Source: https://www.pywhy.org/EconML/_modules/econml/grf/classes.html

Initializes an IVForest estimator with parameters controlling the forest construction, honesty, inference, and computational settings.

```APIDOC
## IVForest.__init__

### Description
Initializes an IVForest estimator with parameters controlling the forest construction, honesty, inference, and computational settings.

### Method
Constructor

### Signature
```python
IVForest(n_estimators=100, *, criterion="mse", max_depth=None, min_samples_split=10, min_samples_leaf=5, min_weight_fraction_leaf=0., min_var_fraction_leaf=None, min_var_leaf_on_val=False, max_features="auto", min_impurity_decrease=0., max_samples=.45, min_balancedness_tol=.45, honest=True, inference=True, fit_intercept=True, subforest_size=4, n_jobs=-1, random_state=None, verbose=0, warm_start=False)
```

### Parameters
- **n_estimators** (int) - Optional - Number of trees in the forest. Default: 100.
- **criterion** (str) - Optional - Split criterion. Default: "mse".
- **max_depth** (int or None) - Optional - Maximum depth of each tree. Default: None.
- **min_samples_split** (int) - Optional - Minimum number of samples required to split an internal node. Default: 10.
- **min_samples_leaf** (int) - Optional - Minimum number of samples required to be at a leaf node. Default: 5.
- **min_weight_fraction_leaf** (float) - Optional - Minimum weighted fraction of the total sum of weights required to be at a leaf node. Default: 0.
- **min_var_fraction_leaf** (float or None) - Optional - Minimum variance fraction required for a leaf. Default: None.
- **min_var_leaf_on_val** (bool) - Optional - Whether to enforce min_var_fraction_leaf on validation set. Default: False.
- **max_features** (str or int) - Optional - Number of features to consider for the best split. Default: "auto".
- **min_impurity_decrease** (float) - Optional - Minimum impurity decrease required for a split. Default: 0.
- **max_samples** (float) - Optional - Fraction of samples to use for each tree. Default: .45.
- **min_balancedness_tol** (float) - Optional - Tolerance for balancedness of splits. Default: .45.
- **honest** (bool) - Optional - Whether to use honest splitting. Default: True.
- **inference** (bool) - Optional - Whether to enable inference. Default: True.
- **fit_intercept** (bool) - Optional - Whether to fit an intercept. Default: True.
- **subforest_size** (int) - Optional - Size of subforests for inference. Default: 4.
- **n_jobs** (int) - Optional - Number of jobs for parallel processing. Default: -1.
- **random_state** (int or None) - Optional - Seed for random number generation. Default: None.
- **verbose** (int) - Optional - Verbosity level. Default: 0.
- **warm_start** (bool) - Optional - Whether to reuse previous fit. Default: False.

### Returns
None
```

--------------------------------

### KernelDML simple example with discrete treatment

Source: https://www.pywhy.org/EconML/_autosummary/econml.dml.KernelDML.html

Shows how to instantiate KernelDML with discrete_treatment=True, dim=10, and bw=5, fit it on synthetic data, and then estimate the effect on the first three samples. The output shows approximate effect values.

```python
from econml.dml import KernelDML

np.random.seed(123)
X = np.random.normal(size=(1000, 5))
T = np.random.binomial(1, scipy.special.expit(X[:, 0]))
y = (1 + .5*X[:, 0]) * T + X[:, 0] + np.random.normal(size=(1000,))
est = KernelDML(discrete_treatment=True, dim=10, bw=5)
est.fit(y, T, X=X, W=None)
```

```python
>>> est.effect(X[:3])
array([0.63042..., 1.86101..., 0.74220...])
```

--------------------------------

### DML example outputs: effect, intervals, coefficients

Source: https://www.pywhy.org/EconML/_autosummary/econml.dml.DML.html

Shows example outputs from calling est.effect, est.effect_interval, est.coef_, est.coef__interval, est.intercept_, and est.intercept__interval after fitting the DML model. Values are truncated with ellipses.

```python
>>> est.effect(X[:3])
array([0.63382..., 1.78225..., 0.71859...])
>>> est.effect_interval(X[:3])
(array([0.27937..., 1.27619..., 0.42091...]),
array([0.98827... , 2.28831..., 1.01628...]))
>>> est.coef_
array([ 0.42857...,  0.04488..., -0.03317...,  0.02258..., -0.14875...])
>>> est.coef__interval()
(array([ 0.25179..., -0.10558..., -0.16723... , -0.11916..., -0.28759...]),
array([ 0.60535...,  0.19536...,  0.10088...,  0.16434..., -0.00990...]))
>>> est.intercept_
1.01166...
>>> est.intercept__interval()
(0.87125..., 1.15207...)
```

--------------------------------

### SparseLinearDRIV.__init__()

Source: https://www.pywhy.org/EconML/

Initializes a SparseLinearDRIV estimator.

```APIDOC
## SparseLinearDRIV.__init__()

### Description
Initializes a SparseLinearDRIV estimator.

### Method
__init__

### Endpoint
SparseLinearDRIV.__init__()

### Parameters
None explicitly documented.

### Returns
None.
```
Tokens:
Raw