Demo for the DoWhy causal API

We show a simple example of adding a causal extension to any dataframe.

import dowhy.datasets
import dowhy.api
from dowhy.graph import build_graph_from_str

import numpy as np
import pandas as pd

from statsmodels.api import OLS

data = dowhy.datasets.linear_dataset(beta=5,
        num_common_causes=1,
        num_instruments = 0,
        num_samples=1000,
        treatment_is_binary=True)
df = data['df']
df['y'] = df['y'] + np.random.normal(size=len(df)) # Adding noise to data. Without noise, the variance in Y|X, Z is zero, and mcmc fails.
nx_graph = build_graph_from_str(data["dot_graph"])

treatment= data["treatment_name"][0]
outcome = data["outcome_name"][0]
common_cause = data["common_causes_names"][0]
df

	W0	v0	y
0	-0.781891	False	0.387390
1	0.870782	False	-1.129183
2	-0.093696	True	7.159128
3	0.351093	True	5.115276
4	1.265067	False	0.173113
...	...	...	...
995	-0.154945	False	-1.722020
996	1.345329	False	0.914354
997	0.580256	True	4.642386
998	0.952156	False	1.785742
999	-0.004864	False	0.524789

1000 rows × 3 columns

# data['df'] is just a regular pandas.DataFrame
df.causal.do(x=treatment,
             variable_types={treatment: 'b', outcome: 'c', common_cause: 'c'},
             outcome=outcome,
             common_causes=[common_cause],
            ).groupby(treatment).mean().plot(y=outcome, kind='bar')

<Axes: xlabel='v0'>

df.causal.do(x={treatment: 1},
              variable_types={treatment:'b', outcome: 'c', common_cause: 'c'},
              outcome=outcome,
              method='weighting',
              common_causes=[common_cause]
              ).groupby(treatment).mean().plot(y=outcome, kind='bar')

<Axes: xlabel='v0'>

cdf_1 = df.causal.do(x={treatment: 1},
              variable_types={treatment: 'b', outcome: 'c', common_cause: 'c'},
              outcome=outcome,
              graph=nx_graph
              )

cdf_0 = df.causal.do(x={treatment: 0},
              variable_types={treatment: 'b', outcome: 'c', common_cause: 'c'},
              outcome=outcome,
              graph=nx_graph
              )

cdf_0

	W0	v0	y	propensity_score	weight
0	-0.188278	False	1.290284	0.483839	2.066804
1	0.028908	False	1.607867	0.472148	2.117982
2	0.556427	False	-0.796434	0.443903	2.252744
3	2.459039	False	0.225063	0.346192	2.888574
4	-0.100968	False	-0.653425	0.479136	2.087090
...	...	...	...	...	...
995	0.910922	False	0.123965	0.425113	2.352315
996	1.041682	False	-0.095624	0.418233	2.391009
997	-0.322524	False	-0.690124	0.491075	2.036349
998	0.303651	False	0.244013	0.457403	2.186254
999	-2.134144	False	-0.111381	0.587866	1.701069

1000 rows × 5 columns

cdf_1

	W0	v0	y	propensity_score	weight
0	-0.218669	True	4.191087	0.514524	1.943545
1	2.934446	True	6.346822	0.676644	1.477883
2	-0.634608	True	5.340087	0.492094	2.032133
3	0.352200	True	5.885853	0.545195	1.834206
4	-0.523283	True	3.163376	0.498098	2.007638
...	...	...	...	...	...
995	1.667235	True	5.699182	0.614198	1.628140
996	0.581155	True	3.079788	0.557414	1.794000
997	-0.388253	True	4.392206	0.505381	1.978707
998	0.984084	True	6.050734	0.578740	1.727892
999	-1.159907	True	6.541946	0.463823	2.155993

1000 rows × 5 columns

Comparing the estimate to Linear Regression

First, estimating the effect using the causal data frame, and the 95% confidence interval.

(cdf_1['y'] - cdf_0['y']).mean()

$\displaystyle 5.06624641268751$

1.96*(cdf_1['y'] - cdf_0['y']).std() / np.sqrt(len(df))

$\displaystyle 0.0927052409804712$

Comparing to the estimate from OLS.

model = OLS(np.asarray(df[outcome]), np.asarray(df[[common_cause, treatment]], dtype=np.float64))
result = model.fit()
result.summary()

OLS Regression Results

Dep. Variable:	y	R-squared (uncentered):	0.936
Model:	OLS	Adj. R-squared (uncentered):	0.936
Method:	Least Squares	F-statistic:	7359.
Date:	Tue, 10 Jun 2025	Prob (F-statistic):	0.00
Time:	16:55:58	Log-Likelihood:	-1415.9
No. Observations:	1000	AIC:	2836.
Df Residuals:	998	BIC:	2846.
Df Model:	2
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
x1	0.3877	0.031	12.569	0.000	0.327	0.448
x2	5.0039	0.045	111.778	0.000	4.916	5.092

Omnibus:	2.260	Durbin-Watson:	1.893
Prob(Omnibus):	0.323	Jarque-Bera (JB):	2.211
Skew:	-0.070	Prob(JB):	0.331
Kurtosis:	2.818	Cond. No.	1.62

Notes:
[1] R² is computed without centering (uncentered) since the model does not contain a constant.
[2] Standard Errors assume that the covariance matrix of the errors is correctly specified.