Quantifying causality in data science with quasi-experiments.

Area of Science:

• Data Science
• Econometrics
• Machine Learning

Background:

• Estimating causality from observational data is crucial but difficult.
• Existing methods often struggle with high-dimensional datasets.

Purpose of the Study:

• To review causal inference methods from econometrics.
• To demonstrate combining quasi-experiments with machine learning for data science.
• To highlight opportunities for data scientists to advance causal estimation.

Main Methods:

• Review of quasi-experimental approaches in econometrics.
• Integration of machine learning techniques with quasi-experiments.
• Application to high-dimensional data challenges.

Main Results:

• Quasi-experiments leverage (quasi) random variation in data.
• Machine learning enhances causal estimation in data science settings.
• Methods can be extended to complex, high-dimensional data.

Conclusions:

• Combining econometrics and machine learning offers powerful causal inference tools.
• Data scientists can play a key role in advancing causal estimation methods.
• These integrated approaches are applicable to medicine, industry, and societal data.