def test_continuous_outcome(self, continuous_data):
g = TimeFixedGFormula(continuous_data,
exposure='A',
outcome='Y',
outcome_type='normal')
g.outcome_model(model='A + W1 + W2 + W3', print_results=False)
g.fit(treatment='all')
npt.assert_allclose(g.marginal_outcome, -0.730375, rtol=1e-5)
def test_warn_categorical_treatment(self, cat_data):
g = TimeFixedGFormula(cat_data,
exposure=['A1', 'A2'],
exposure_type='categorical',
outcome='Y')
g.outcome_model(model='A1 + A2', print_results=False)
with pytest.warns(UserWarning):
g.fit(treatment=['True', 'True'])
def test_g_formula1(self, sim_t_fixed_data):
g = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
g.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
g.fit(treatment='all')
r1 = g.marginal_outcome
g.fit(treatment='none')
r0 = g.marginal_outcome
npt.assert_allclose(r1 - r0, -0.075186, rtol=1e-2)
def test_error_mismatch_categorical_treatment(self, cat_data):
g = TimeFixedGFormula(cat_data,
exposure=['A1', 'A2'],
exposure_type='categorical',
outcome='Y')
g.outcome_model(model='A1 + A2', print_results=False)
with pytest.raises(ValueError):
g.fit(treatment=['True', 'False', 'False'])
def test_conditional_stochastic_warning(self):
data = pd.DataFrame()
data['A'] = [1]*50 + [0]*50
data['L'] = [1]*25 + [0]*25 + [1]*40 + [0]*10
data['Y'] = [1]*25 + [0]*25 + [1]*25 + [0]*25
g = TimeFixedGFormula(data, exposure='A', outcome='Y')
g.outcome_model(model='A + L + A:L', print_results=False)
with pytest.warns(UserWarning):
g.fit_stochastic(p=[1.0, 1.0], conditional=["(g['L']==1) | (g['L']==0)", "g['L']==0"])
def test_weighted_data(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y', weights='w')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
r1 = g.marginal_outcome
npt.assert_allclose(r1, 0.564286, rtol=1e-5)
g.fit(treatment='none')
r0 = g.marginal_outcome
npt.assert_allclose(r0, 0.404286, rtol=1e-5)
def test_categorical_treat(self, cat_data):
g = TimeFixedGFormula(cat_data, exposure=['A1', 'A2'], exposure_type='categorical', outcome='Y')
g.outcome_model(model='A1 + A2', print_results=False)
g.fit(treatment=["False", "False"])
npt.assert_allclose(g.marginal_outcome, 0.373091, rtol=1e-5)
g.fit(treatment=["True", "False"])
npt.assert_allclose(g.marginal_outcome, 0.8128, rtol=1e-5)
g.fit(treatment=["False", "True"])
npt.assert_allclose(g.marginal_outcome, 0.5025, rtol=1e-5)
def test_g_formula2(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
r1 = g.marginal_outcome
npt.assert_allclose(r1, 0.575)
g.fit(treatment='none')
r0 = g.marginal_outcome
npt.assert_allclose(r0, 0.415)
def test_weight_stochastic_between_marginal(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y', weights='w')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
r1 = g.marginal_outcome
g.fit(treatment='none')
r0 = g.marginal_outcome
g.fit_stochastic(p=0.5)
r_star = g.marginal_outcome
assert r0 < r_star < r1
def test_directions_correct(self, sim_t_fixed_data):
g = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
g.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
g.fit(treatment='all')
r1 = g.marginal_outcome
g.fit(treatment='none')
r0 = g.marginal_outcome
g.fit(treatment="g['W3'] > 2")
rc = g.marginal_outcome
assert r1 < rc < r0
def test_iterative_for_single_t(self, sim_t_fixed_data):
# Estimating sequential regression for single t
gt = IterativeCondGFormula(sim_t_fixed_data, exposures=['A'], outcomes=['Y'])
gt.outcome_model(['A + W1_sq + W2 + W3'], print_results=False)
gt.fit(treatments=[1])
# Estimating with TimeFixedGFormula
gf = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
gf.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
gf.fit(treatment='all')
# Expected behavior; same results between the estimation methods
npt.assert_allclose(gf.marginal_outcome, gt.marginal_outcome)
def test_monte_carlo_for_single_t(self, sim_t_fixed_data):
# Estimating monte carlo for single t
gt = MonteCarloGFormula(sim_t_fixed_data, idvar='id', exposure='A', outcome='Y', time_out='t', time_in='t0')
gt.outcome_model('A + W1_sq + W2 + W3', print_results=False)
gt.exposure_model('W1_sq', print_results=False)
gt.fit(treatment="all", sample=1000000) # Keep this a high number to reduce simulation errors
print(gt.predicted_outcomes)
# Estimating with TimeFixedGFormula
gf = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
gf.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
gf.fit(treatment='all')
# Expected behavior; same results between the estimation methods
npt.assert_allclose(gf.marginal_outcome, np.mean(gt.predicted_outcomes['Y']), rtol=1e-3)
def test_weight_stochastic_matches_marginal(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y', weights='w')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
rm = g.marginal_outcome
g.fit_stochastic(p=1.0)
rs = g.marginal_outcome
npt.assert_allclose(rm, rs, rtol=1e-7)
g.fit(treatment='none')
rm = g.marginal_outcome
g.fit_stochastic(p=0.0)
rs = g.marginal_outcome
npt.assert_allclose(rm, rs, rtol=1e-7)
def test_conditional_weight_stochastic_matches_marginal(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y', weights='w')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
rm = g.marginal_outcome
g.fit_stochastic(p=[1.0, 1.0], conditional=["g['L']==1", "g['L']==0"])
rs = g.marginal_outcome
npt.assert_allclose(rm, rs, rtol=1e-7)
g.fit(treatment='none')
rn = g.marginal_outcome
g.fit_stochastic(p=[0.0, 0.0], conditional=["g['L']==1", "g['L']==0"])
rs = g.marginal_outcome
npt.assert_allclose(rn, rs, rtol=1e-7)
def test_exposed_standardized1(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y', standardize='exposed')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
rm = g.marginal_outcome
g.fit(treatment='none')
rs = g.marginal_outcome
npt.assert_allclose(rm - rs, 0.16, rtol=1e-5)
npt.assert_allclose(rm / rs, 1.387097, rtol=1e-5)
g = TimeFixedGFormula(data, exposure='A', outcome='Y', standardize='unexposed')
g.outcome_model(model='A + L + A:L', print_results=False)
g.fit(treatment='all')
rm = g.marginal_outcome
g.fit(treatment='none')
rs = g.marginal_outcome
npt.assert_allclose(rm - rs, 0.16, rtol=1e-5)
npt.assert_allclose(rm / rs, 1.384615, rtol=1e-5)
def test_sequential_regression_for_single_t(self, sim_t_fixed_data):
# Estimating sequential regression for single t
gt = TimeVaryGFormula(sim_t_fixed_data,
idvar='id',
exposure='A',
outcome='Y',
time_out='t',
method='SequentialRegression')
gt.outcome_model('A + W1_sq + W2 + W3', print_results=False)
gt.fit(treatment="all")
# Estimating with TimeFixedGFormula
gf = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
gf.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
gf.fit(treatment='all')
# Expected behavior; same results between the estimation methods
npt.assert_allclose(gf.marginal_outcome, gt.predicted_outcomes)
def causal_check():
data = load_sample_data(False).drop(columns=['cd4_wk45'])
data[['cd4_rs1', 'cd4_rs2']] = spline(data,
'cd40',
n_knots=3,
term=2,
restricted=True)
data[['age_rs1', 'age_rs2']] = spline(data,
'age0',
n_knots=3,
term=2,
restricted=True)
# Check TimeFixedGFormula diagnostics
g = TimeFixedGFormula(data, exposure='art', outcome='dead')
g.outcome_model(
model=
'art + male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0'
)
g.run_diagnostics(decimal=3)
# Check IPTW plots
ipt = IPTW(data, treatment='art', outcome='dead')
ipt.treatment_model(
'male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0',
stabilized=True)
ipt.marginal_structural_model('art')
ipt.fit()
ipt.plot_love()
plt.tight_layout()
plt.show()
ipt.plot_kde()
plt.show()
ipt.plot_kde(measure='logit')
plt.show()
ipt.plot_boxplot()
plt.show()
ipt.plot_boxplot(measure='logit')
plt.show()
ipt.run_diagnostics()
# Check AIPTW Diagnostics
aipw = AIPTW(data, exposure='art', outcome='dead')
aipw.exposure_model(
'male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0')
aipw.outcome_model(
'art + male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0'
)
aipw.fit()
aipw.run_diagnostics()
aipw.plot_kde(to_plot='exposure')
plt.show()
aipw.plot_kde(to_plot='outcome')
plt.show()
aipw.plot_love()
plt.show()
# Check TMLE diagnostics
tmle = TMLE(data, exposure='art', outcome='dead')
tmle.exposure_model(
'male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0')
tmle.outcome_model(
'art + male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0'
)
tmle.fit()
tmle.run_diagnostics()
tmle.plot_kde(to_plot='exposure')
plt.show()
tmle.plot_kde(to_plot='outcome')
plt.show()
tmle.plot_love()
plt.show()
# Check SurvivalGFormula plots
df = load_sample_data(False).drop(columns=['cd4_wk45'])
df['t'] = np.round(df['t']).astype(int)
df = pd.DataFrame(np.repeat(df.values, df['t'], axis=0),
columns=df.columns)
df['t'] = df.groupby('id')['t'].cumcount() + 1
df.loc[((df['dead'] == 1) & (df['id'] != df['id'].shift(-1))), 'd'] = 1
df['d'] = df['d'].fillna(0)
df['t_sq'] = df['t']**2
df['t_cu'] = df['t']**3
sgf = SurvivalGFormula(df,
idvar='id',
exposure='art',
outcome='d',
time='t')
sgf.outcome_model(
model='art + male + age0 + cd40 + dvl0 + t + t_sq + t_cu')
sgf.fit(treatment='all')
sgf.plot()
plt.show()
sgf.plot(c='r', linewidth=3, alpha=0.8)
plt.show()
#Crude Model
ze.RiskRatio(df, exposure='art', outcome='dead')
ze.RiskDiff(df, exposure='art', outcome='dead')
#Adjusted Model
model = 'art + male + age0 + cd40 + dvl0'
f = sm.families.family.Binomial(sm.families.links.identity)
linrisk = smf.glm('dead ~ ' + model, df, family=f).fit()
linrisk.summary()
f = sm.families.family.Binomial(sm.families.links.log)
log = smf.glm('dead ~ art', df, family=f).fit()
log.summary()
#g-formula
g = TimeFixedGFormula(df, exposure='art', outcome='dead')
g.outcome_model(
model=
'art + male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0')
g.fit(treatment='all')
r_all = g.marginal_outcome
g.fit(treatment='none')
r_none = g.marginal_outcome
print('RD1 = ', r_all - r_none)
print('RR1 = ', r_all / r_none)
rd_results = []
rr_results = []
for i in range(500):
dfs = df.sample(n=df.shape[0], replace=True)
g = TimeFixedGFormula(dfs, exposure='art', outcome='dead')
g.outcome_model(
model=
'art + male + age0 + age_rs1 + age_rs2 + cd40 + cd4_rs1 + cd4_rs2 + dvl0',
def test_custom_treatment(self, sim_t_fixed_data):
g = TimeFixedGFormula(sim_t_fixed_data, exposure='A', outcome='Y')
g.outcome_model(model='A + W1_sq + W2 + W3', print_results=False)
g.fit(treatment="g['W3'] > 2")
npt.assert_allclose(g.marginal_outcome, 0.682829, rtol=1e-5)
def test_error_binary_exposure_list_treatments(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y')
g.outcome_model(model='A + L + A:L', print_results=False)
with pytest.raises(ValueError):
g.fit(treatment=['True', 'False'])
def test_stochastic_conditional_probability(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y')
g.outcome_model(model='A + L + A:L', print_results=False)
with pytest.raises(ValueError):
g.fit_stochastic(p=[0.0], conditional=["g['L']==1", "g['L']==0"])
def test_error_wrong_treatment_object(self, data):
g = TimeFixedGFormula(data, exposure='A', outcome='Y')
g.outcome_model(model='A + L + A:L', print_results=False)
with pytest.raises(ValueError):
g.fit(treatment=5)
