import pandas as pd
import numpy as np
np.random.seed(24)
n = 5000
email = np.random.binomial(1, 0.5, n)
credit_limit = np.random.gamma(6, 200, n)
risk_score = np.random.beta(credit_limit, credit_limit.mean(), n)
opened = np.random.normal(5 + 0.001*credit_limit - 4*risk_score, 2)
opened = (opened > 4).astype(float) * email
agreement = np.random.normal(30 +(-0.003*credit_limit - 10*risk_score), 7) * 2 * opened
agreement = (agreement > 40).astype(float)
payments = (np.random.normal(500 + 0.16*credit_limit - 40*risk_score + 11*agreement + email, 75).astype(int) // 10) * 10
data = pd.DataFrame(dict(payments=payments,
email=email,
opened=opened,
agreement=agreement,
credit_limit=credit_limit,
risk_score=risk_score))
data.to_csv("collections_email.csv", index=False)
import pandas as pd
import numpy as np
np.random.seed(24)
n = 80
hospital = np.random.binomial(1, 0.5, n)
treatment = np.where(hospital.astype(bool),
np.random.binomial(1, 0.9, n),
np.random.binomial(1, 0.1, n))
severity = np.where(hospital.astype(bool),
np.random.normal(20, 5, n),
np.random.normal(10, 5, n))
days = np.random.normal(15 + -5*treatment + 2*severity, 7).astype(int)
hospital = pd.DataFrame(dict(hospital=hospital,
treatment=treatment,
severity=severity,
days=days))
hospital.to_csv("hospital_treatment.csv", index=False)
import pandas as pd
import numpy as np
np.random.seed(24)
n = 10000
push_assigned = np.random.binomial(1, 0.5, n)
income = np.random.gamma(6, 200, n)
push_delivered = np.random.normal(5 + 0.3+income, 500)
push_delivered = ((push_delivered > 800) & (push_assigned == 1)).astype(int)
in_app_purchase = (np.random.normal(100 + 20*push_delivered + 0.5*income, 75).astype(int) // 10)
data = pd.DataFrame(dict(in_app_purchase=in_app_purchase,
push_assigned=push_assigned,
push_delivered=push_delivered))
data.to_csv("app_engagement_push.csv", index=False)
import numpy as np
import pandas as pd
def make_confounded_data(N):
def get_severity(df):
return ((np.random.beta(1, 3, size=df.shape[0]) * (df["age"] < 30)) +
(np.random.beta(3, 1.5, size=df.shape[0]) * (df["age"] >= 30)))
def get_treatment(df):
return ((.33 * df["sex"] +
1.5 * df["severity"] + df["severity"] ** 2 +
0.15 * np.random.normal(size=df.shape[0])) > 1.5).astype(int)
def get_recovery(df):
return ((2 +
0.5 * df["sex"] +
0.03 * df["age"] + 0.03 * ((df["age"] * 0.1) ** 2) +
df["severity"] + np.log(df["severity"]) +
df["sex"] * df["severity"] -
df["medication"]) * 10).astype(int)
np.random.seed(1111)
sexes = np.random.randint(0, 2, size=N)
ages = np.random.gamma(8, scale=4, size=N)
meds = np.random.beta(1, 1, size=N)
# dados com designação aleatória
df_rnd = pd.DataFrame(dict(sex=sexes, age=ages, medication=meds))
df_rnd['severity'] = get_severity(df_rnd)
df_rnd['recovery'] = get_recovery(df_rnd)
features = ['sex', 'age', 'severity', 'medication', 'recovery']
df_rnd = df_rnd[features] # to enforce column order
# dados observacionais
df_obs = df_rnd.copy()
df_obs['medication'] = get_treatment(df_obs)
df_obs['recovery'] = get_recovery(df_obs)
# dados contrafactuais data
df_ctf = df_obs.copy()
df_ctf['medication'] = ((df_ctf['medication'] == 1) ^ 1).astype(float)
df_ctf['recovery'] = get_recovery(df_ctf)
return df_rnd, df_obs, df_ctf
np.random.seed(1234)
df_rnd, df_obs, df_ctf = make_confounded_data(20000)
df_obs.to_csv("medicine_impact_recovery.csv", index=False)
import pandas as pd
import numpy as np
np.random.seed(123)
POAMay = np.random.gamma(7,10, 500) * np.random.binomial(1, .7, 500)
POAJul = np.random.gamma(7,15, 800) * np.random.binomial(1, .8, 800)
FLMay = np.random.gamma(10,20, 1300) * np.random.binomial(1, .85, 1300)
FLJul = np.random.gamma(11,21, 2000) * np.random.binomial(1, .9, 2000)
data = pd.concat([
pd.DataFrame(dict(deposits = POAMay.astype(int), poa=1, jul=0)),
pd.DataFrame(dict(deposits = POAJul.astype(int), poa=1, jul=1)),
pd.DataFrame(dict(deposits = FLMay.astype(int), poa=0, jul=0)),
pd.DataFrame(dict(deposits = FLJul.astype(int), poa=0, jul=1))
])
data.to_csv("billboard_impact.csv", index=False)