def test_summary():
Y = np.array([1, 2, 3, 4, 6, 5])
D = np.array([0, 0, 1, 1, 0, 1])
X = np.array([[1, 3], [5, 7], [8, 6], [4, 2], [9, 11], [12, 10]])
data = d.Data(Y, D, X)
summary = s.Summary(data)
N = 6
K = 2
N_c = 3
N_t = 3
Y_c_mean = 3
Y_t_mean = 4
Y_c_sd = np.sqrt(7)
Y_t_sd = 1
rdiff = 1
X_c_mean = np.array([5, 7])
X_t_mean = np.array([8, 6])
X_c_sd = np.array([4, 4])
X_t_sd = np.array([4, 4])
ndiff = np.array([0.75, -0.25])
keys1 = {
'N', 'K', 'N_c', 'N_t', 'Y_c_mean', 'Y_t_mean', 'Y_c_sd', 'Y_t_sd',
'X_c_mean', 'X_t_mean', 'X_c_sd', 'X_t_sd', 'rdiff', 'ndiff',
'sum_of_abs_ndiffs', 'num_large_ndiffs', 'num_outputs'
}
assert_equal(summary['N'], N)
assert_equal(summary['N_c'], N_c)
assert_equal(summary['N_t'], N_t)
assert_equal(summary['Y_c_mean'], Y_c_mean)
assert_equal(summary['Y_t_mean'], Y_t_mean)
assert_equal(summary['Y_c_sd'], Y_c_sd)
assert_equal(summary['Y_t_sd'], Y_t_sd)
assert_equal(summary['rdiff'], rdiff)
assert np.array_equal(summary['X_c_mean'], X_c_mean)
assert np.array_equal(summary['X_t_mean'], X_t_mean)
assert np.array_equal(summary['X_c_sd'], X_c_sd)
assert np.array_equal(summary['X_t_sd'], X_t_sd)
assert np.array_equal(summary['ndiff'], ndiff)
assert_equal(set(summary.keys()), keys1)
p_c = np.array([0.3, 0.5, 0.7])
p_t = np.array([0.1, 0.5, 0.9])
summary._summarize_pscore(p_c, p_t)
keys2 = {
'N', 'K', 'N_c', 'N_t', 'Y_c_mean', 'Y_t_mean', 'Y_c_sd', 'Y_t_sd',
'X_c_mean', 'X_t_mean', 'X_c_sd', 'X_t_sd', 'rdiff', 'ndiff', 'p_min',
'p_max', 'p_c_mean', 'p_t_mean', 'sum_of_abs_ndiffs',
'num_large_ndiffs', 'num_outputs'
}
assert_equal(summary['p_min'], 0.1)
assert_equal(summary['p_max'], 0.9)
assert_equal(summary['p_c_mean'], 0.5)
assert_equal(summary['p_t_mean'], 0.5)
assert_equal(set(summary.keys()), keys2)
def test_propensity():
import causalinference.core.data as d
import causalinference.core.propensity as p
from utils import random_data
D = np.array([0, 0, 0, 1, 1, 1])
X = np.array([[7, 8], [3, 10], [7, 10], [4, 7], [5, 10], [9, 8]])
Y = random_data(D_cur=D, X_cur=X)
print Y
data = d.Data(Y, D, X)
propensity = p.Propensity(data, [0, 1], [])
print propensity
def test_weighting():
Y = np.array([1, -2, 3, -5, 7])
D = np.array([0, 1, 0, 1, 0])
X = np.array([3, 2, 3, 5, 5])
pscore = np.array([0.1, 0.25, 0.5, 0.75, 0.9])
data = d.Data(Y, D, X)
data._dict['pscore'] = pscore
weighting = w.Weighting(data)
ate = -6.7963178
ate_se = 2.8125913
keys = {'ate', 'ate_se'}
assert np.allclose(weighting['ate'], ate)
assert np.allclose(weighting['ate_se'], ate_se)
assert_equal(set(weighting.keys()), keys)
def test_ols():
Y = np.array([52, 30, 5, 29, 12, 10, 44, 87])
D = np.array([0, 0, 0, 0, 1, 1, 1, 1])
X = np.array([[1, 42], [3, 32], [9, 7], [12, 86],
[5, 94], [4, 36], [2, 13], [6, 61]])
data = d.Data(Y, D, X)
adj1 = 0
ols1 = o.OLS(data, adj1, [])
ate1 = 9.25
ate_se1 = 17.68253
keys = {'ate', 'ate_se',
'atc',
'att',
'atc_se',
'name_to_coef',
'att_se',
'r2'
}
assert np.allclose(ols1['ate'], ate1)
assert np.allclose(ols1['ate_se'], ate_se1)
assert_equal(set(ols1.keys()), keys)
adj2 = 1
ols2 = o.OLS(data, adj2, [])
ate2 = 3.654552
ate_se2 = 17.749993
assert np.allclose(ols2['ate'], ate2)
assert np.allclose(ols2['ate_se'], ate_se2)
assert_equal(set(ols2.keys()), keys)
adj3 = 2
ols3 = o.OLS(data, adj3, feature_names=[])
ate3 = 30.59444
atc3 = 63.2095
att3 = -2.020611
ate_se3 = 19.91887865
atc_se3 = 29.92152
att_se3 = 11.8586
assert np.allclose(ols3['ate'], ate3)
assert np.allclose(ols3['atc'], atc3)
assert np.allclose(ols3['att'], att3)
assert np.allclose(ols3['ate_se'], ate_se3)
assert np.allclose(ols3['atc_se'], atc_se3)
assert np.allclose(ols3['att_se'], att_se3)
assert_equal(set(ols3.keys()), keys)
def test_propensityselect():
D = np.array([0, 0, 0, 1, 1, 1])
X = np.array([[7, 8], [3, 10], [7, 10], [4, 7], [5, 10], [9, 8]])
Y = random_data(D_cur=D, X_cur=X)
data = d.Data(Y, D, X)
propensity1 = p.PropensitySelect(data, [], 1, 2.71)
lin1 = [1]
qua1 = []
coef1 = np.array([6.5424027, -0.7392041])
loglike1 = -3.627939
fitted1 = np.array(
[0.6522105, 0.2995088, 0.2995088, 0.7970526, 0.2995088, 0.6522105])
se1 = np.array([6.8455179, 0.7641445])
keys = {'lin', 'qua', 'coef', 'loglike', 'fitted', 'se'}
assert_equal(propensity1['lin'], lin1)
assert_equal(propensity1['qua'], qua1)
assert np.allclose(propensity1['coef'], coef1)
assert np.allclose(propensity1['loglike'], loglike1)
assert np.allclose(propensity1['fitted'], fitted1)
assert np.allclose(propensity1['se'], se1)
assert_equal(set(propensity1.keys()), keys)
propensity2 = p.PropensitySelect(data, [0, 1], 1, 2.71)
lin2 = [0, 1]
qua2 = []
coef2 = np.array([6.8066090, -0.0244874, -0.7524939])
loglike2 = -3.626517
fitted2 = np.array(
[0.6491366, 0.3117840, 0.2911631, 0.8086407, 0.3013733, 0.6379023])
se2 = np.array([8.5373779, 0.4595191, 0.8106499])
assert_equal(propensity2['lin'], lin2)
assert_equal(propensity2['qua'], qua2)
assert np.allclose(propensity2['coef'], coef2)
assert np.allclose(propensity2['loglike'], loglike2)
assert np.allclose(propensity2['fitted'], fitted2)
assert np.allclose(propensity2['se'], se2)
def test_propensity():
D = np.array([0, 0, 0, 1, 1, 1])
X = np.array([[7, 8], [3, 10], [7, 10], [4, 7], [5, 10], [9, 8]])
Y = random_data(D_cur=D, X_cur=X)
data = d.Data(Y, D, X)
propensity = p.Propensity(data, [0, 1], [])
lin = [0, 1]
qua = []
coef = np.array([6.8066090, -0.0244874, -0.7524939])
loglike = -3.626517
fitted = np.array([0.6491366, 0.3117840, 0.2911631,
0.8086407, 0.3013733, 0.6379023])
se = np.array([8.5373779, 0.4595191, 0.8106499])
keys = {'lin', 'qua', 'coef', 'loglike', 'fitted', 'se'}
assert_equal(propensity['lin'], lin)
assert_equal(propensity['qua'], qua)
assert np.allclose(propensity['coef'], coef)
assert np.allclose(propensity['loglike'], loglike)
assert np.allclose(propensity['fitted'], fitted)
assert np.allclose(propensity['se'], se)
assert_equal(set(propensity.keys()), keys)
def test_data():
Y1 = np.array([1.2, 3.45, -6, 78.90, -9, 8.7654])
D1 = np.array([0, 1, 0, 1.0, 0.0, 1])
X1 = np.array([[-1, 2], [3, -4], [-5.6, -7], [8.9, 0.0], [99, 877],
[-666, 54321]])
data = d.Data(Y1, D1, X1)
ans1 = np.array([[1.2, 3.45, -6, 78.9, -9, 8.7654]]).transpose()
print(ans1)
print(data['Y'])
assert np.array_equal(data['Y'], ans1)
ans2 = np.array([0, 1, 0, 1, 0, 1])
assert np.array_equal(data['D'], ans2)
ans3 = np.array([[-1, 2], [3, -4], [-5.6, -7], [8.9, 0], [99, 877],
[-666, 54321]])
assert np.array_equal(data['X'], ans3)
ans4 = 6
assert_equal(data['N'], ans4)
ans5 = 2
assert_equal(data['K'], ans5)
ans6 = np.array([True, False, True, False, True, False])
assert np.array_equal(data['controls'], ans6)
ans7 = np.array([False, True, False, True, False, True])
assert np.array_equal(data['treated'], ans7)
ans8 = np.array([[1.2, -6, -9]]).transpose()
assert np.array_equal(data['Y_c'], ans8)
ans9 = np.array([[3.45, 78.9, 8.7654]]).transpose()
assert np.array_equal(data['Y_t'], ans9)
ans10 = np.array([[-1, 2], [-5.6, -7], [99, 877]])
assert np.array_equal(data['X_c'], ans10)
ans11 = np.array([[3, -4], [8.9, 0], [-666, 54321]])
assert np.array_equal(data['X_t'], ans11)
ans12 = 3
assert_equal(data['N_t'], ans12)
ans13 = 3
assert_equal(data['N_c'], ans13)
ans14 = 'int'
assert_equal(data['D'].dtype, ans14)
ans15 = {
'Y', 'D', 'X', 'N', 'K', 'controls', 'treated', 'Y_c', 'Y_t', 'X_c',
'X_t', 'N_c', 'N_t', 'pairs', 'ids', 'num_outputs'
}
assert_equal(set(data.keys()), ans15)
Y2 = np.array([[1.2], [3.45], [-6], [78.90]])
D2 = np.array([[0], [1], [0.0], [1]])
X2 = np.array([[-1, 2], [3, -4], [-5.6, -7], [8.9, 0.0]])
assert_raises(ValueError, d.Data, Y2, D2, X2)