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	0.692489	True	6.513994
1	0.647121	True	8.080264
2	-0.291692	True	3.450810
3	1.054548	True	8.130324
4	-0.618836	True	3.947363
...	...	...	...
995	2.401029	True	9.171075
996	1.724637	True	8.373037
997	0.999508	True	7.068687
998	-0.378417	False	1.271395
999	-0.114634	True	4.968926

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	1.232935	False	0.530949	0.070514	14.181545
1	-0.120800	False	-1.078003	0.539004	1.855275
2	0.601958	False	2.265526	0.213499	4.683865
3	0.292528	False	0.574488	0.336535	2.971462
4	1.000027	False	0.634646	0.108299	9.233726
...	...	...	...	...	...
995	1.312027	False	3.808483	0.060733	16.465611
996	0.707995	False	1.570884	0.179726	5.564022
997	0.805173	False	1.552603	0.152575	6.554158
998	0.598637	False	-0.046964	0.214628	4.659231
999	0.204984	False	-1.838001	0.377095	2.651853

1000 rows × 5 columns

cdf_1

	W0	v0	y	propensity_score	weight
0	0.411220	True	5.997172	0.714754	1.399083
1	1.470005	True	7.939459	0.955118	1.046991
2	0.302906	True	6.055634	0.668131	1.496713
3	0.763221	True	5.912023	0.836140	1.195972
4	-0.505381	True	3.969568	0.282244	3.543032
...	...	...	...	...	...
995	0.879454	True	5.608546	0.865835	1.154954
996	2.503162	True	10.771813	0.994207	1.005827
997	2.256445	True	8.361526	0.990499	1.009593
998	1.252414	True	6.828937	0.932022	1.072936
999	1.605996	True	8.848931	0.965529	1.035701

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()

$\displaystyle 5.60782363962144$

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

$\displaystyle 0.16460849994063$

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):	0.973
Model:	OLS	Adj. R-squared (uncentered):	0.973
Method:	Least Squares	F-statistic:	1.808e+04
Date:	Tue, 31 Dec 2024	Prob (F-statistic):	0.00
Time:	08:35:02	Log-Likelihood:	-1451.6
No. Observations:	1000	AIC:	2907.
Df Residuals:	998	BIC:	2917.
Df Model:	2
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
x1	1.9217	0.040	48.081	0.000	1.843	2.000
x2	5.0082	0.058	86.916	0.000	4.895	5.121

Omnibus:	0.132	Durbin-Watson:	1.898
Prob(Omnibus):	0.936	Jarque-Bera (JB):	0.193
Skew:	0.020	Prob(JB):	0.908
Kurtosis:	2.945	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.