def linear_data_pandas():
w, t, y = generate_wty_linear_multi_w_data(n=N,
wdim=5,
binary_treatment=True,
delta=ATE,
data_format='pandas')
return w, t, y
def linear_gen_model():
w, t, y = generate_wty_linear_multi_w_data(n=N,
wdim=5,
binary_treatment=True,
delta=ATE,
data_format='numpy')
return LinearGenModel(w, t, y, binary_treatment=True)
def test_ipw_numpy_data():
w, t, y = generate_wty_linear_multi_w_data(n=N,
wdim=5,
binary_treatment=True,
delta=ATE,
data_format='numpy')
ipw = IPWEstimator()
ipw.fit(w, t, y)
assert ipw.estimate_ate() == approx(ATE, rel=.2)
return t_samples
def _sample_y(self, t, w=None):
y_samples = self.linear_gaussian_sampler(np.concatenate([w, t], 1),
self.beta_y_tw,
self.sigma_y_tw)
return y_samples
def mean_y(self, t, w):
X = np.concatenate([w, t], 1)
return self._pad_with_ones(X).dot(self.beta_y_tw)
if __name__ == '__main__':
from data.synthetic import generate_wty_linear_multi_w_data
from utils import NUMPY
# data = generate_wty_linear_multi_w_data(500, data_format=NUMPY, wdim=5)
#
# dgm = DataGenModel(data)
# data_samples = dgm.sample()
w, t, y = generate_wty_linear_multi_w_data(500, data_format=NUMPY, wdim=5)
lgm = LinearGenModel(w, t, y)
data_samples = lgm.sample()
lgm.plot_ty_dists()
uni_metrics = lgm.get_univariate_quant_metrics()
multi_ty_metrics = lgm.get_multivariate_quant_metrics(include_w=False)
multi_wty_metrics = lgm.get_multivariate_quant_metrics(include_w=True)
def linear_data():
w, t, y = generate_wty_linear_multi_w_data(n=N,
wdim=5,
binary_treatment=True,
delta=ATE)
return w, t, y
def linear_gen_model():
w, t, y = generate_wty_linear_multi_w_data(N,
data_format='numpy',
wdim=5,
delta=ATE)
return LinearGenModel(w, t, y)
def linear_gen_model_train_test():
w, t, y = generate_wty_linear_multi_w_data(N,
data_format='numpy',
wdim=5,
delta=ATE)
return LinearGenModel(w, t, y, train_prop=.5, test_prop=.5)
def test_multivariate_w_data():
n = 10
d = 5
w, t, y = generate_wty_linear_multi_w_data(n, wdim=5)
assert all(isinstance(x, np.ndarray) for x in (w, t, y))
assert w.shape == (n, d) and t.shape == (n, ) and y.shape == (n, )
match_out=match_out,
distance_tolerance=distance_tolerance,
tolerance=tolerance,
version=version)
def gen_match(Tr, X, estimand='ATT'):
# Install the Matching and rgenoud R packages, if not already installed
to_install = [
x for x in [MATCHING, RGENOUD] if not rpackages.isinstalled(x)
]
if len(to_install) > 0:
utils = importr('utils')
utils.chooseCRANmirror(ind=1) # select the first mirror in the list
utils.install_packages(StrVector(to_install))
matching = importr(MATCHING) # import Matching R package
importr(RGENOUD)
return matching.GenMatch(Tr=Tr, X=X, estimand=estimand)
if __name__ == '__main__':
w, t, y = generate_wty_linear_multi_w_data(100,
wdim=5,
binary_treatment=True,
delta=5)
m = MatchingEstimator(weighting=GENETIC)
m.fit(w, t, y)
est = m.estimate_ate()
conf_int = m.ate_conf_int()