Demo for the DoWhy causal API

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

import dowhy
dowhy.enable_notebook_rendering()

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	1.946486	True	9.601481
1	0.830639	False	1.848806
2	-0.146411	True	6.541884
3	1.111248	True	6.600895
4	-0.059885	True	3.016531
...	...	...	...
995	0.091708	True	4.413204
996	0.632677	True	4.706617
997	0.437943	True	5.230384
998	2.558886	True	9.743775
999	1.665350	True	8.772056

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	0.397159	False	0.659022	0.369388	2.707182
1	2.229516	False	4.665517	0.046024	21.727891
2	0.463989	False	1.447101	0.348441	2.869930
3	1.154969	False	1.776468	0.172590	5.794071
4	-0.686871	False	-2.221900	0.719539	1.389778
...	...	...	...	...	...
995	0.651798	False	0.020120	0.292806	3.415234
996	1.505612	False	1.669227	0.114544	8.730250
997	-1.652279	False	-3.144685	0.905302	1.104604
998	0.374365	False	1.510566	0.376651	2.654976
999	1.025267	False	1.324563	0.199303	5.017491

1000 rows × 5 columns

cdf_1

	W0	v0	y	propensity_score	weight
0	0.007914	True	5.214271	0.501122	1.995523
1	1.886673	True	8.859618	0.928531	1.076969
2	0.988608	True	6.329401	0.792606	1.261660
3	0.801659	True	6.300524	0.747625	1.337569
4	0.382875	True	6.355585	0.626067	1.597273
...	...	...	...	...	...
995	1.167872	True	6.102895	0.829906	1.204956
996	1.473154	True	7.980661	0.880893	1.135211
997	0.553763	True	6.335199	0.678791	1.473208
998	0.259339	True	6.149736	0.585905	1.706763
999	1.017072	True	6.894531	0.798909	1.251707

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()

$\displaystyle 5.13983696214543$

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

$\displaystyle 0.172382321522194$

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):	0.976
Model:	OLS	Adj. R-squared (uncentered):	0.976
Method:	Least Squares	F-statistic:	2.047e+04
Date:	Sun, 19 Jul 2026	Prob (F-statistic):	0.00
Time:	05:31:53	Log-Likelihood:	-1370.0
No. Observations:	1000	AIC:	2744.
Df Residuals:	998	BIC:	2754.
Df Model:	2
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
x1	1.8546	0.035	53.162	0.000	1.786	1.923
x2	4.9893	0.053	93.914	0.000	4.885	5.094

Omnibus:	2.229	Durbin-Watson:	1.924
Prob(Omnibus):	0.328	Jarque-Bera (JB):	2.284
Skew:	0.112	Prob(JB):	0.319
Kurtosis:	2.933	Cond. No.	2.85

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.