Causal inference in practice: Here, there, causality is everywhere

25
Here, there, causality is everywhere Amit Sharma Microsoft Research, New York

Transcript of Causal inference in practice: Here, there, causality is everywhere

Page 1: Causal inference in practice: Here, there, causality is everywhere

Here, there, causality is everywhereAmit SharmaMicrosoft Research, New York

Page 2: Causal inference in practice: Here, there, causality is everywhere

My route to causality

Building recommender

systems in social Networks

Conducting user

experiments

Estimating impact of

recommendations and social

feeds

Page 3: Causal inference in practice: Here, there, causality is everywhere

Causality is everywhere Spans every branch of science. Aristotle: to know, is to know the final cause.

Two approaches: Randomized experiments (Fisher): Gold standard Observational data: Messy.

Page 4: Causal inference in practice: Here, there, causality is everywhere

Outline Causality is everywhere Economics Political Science Human Behavior Biology and Medicine Online systems Estimating causality using graphical models Conditioning Mechanism-based Natural Experiments The promise of graphical models

Page 5: Causal inference in practice: Here, there, causality is everywhere

Causality in economics

David Card. The causal effect of education on earnings (1999)

Conley and Heerwig. The Long-Term Effects of Military Conscription on Mortality: Estimates From the Vietnam-Era Draft Lottery (2012)

Page 6: Causal inference in practice: Here, there, causality is everywhere

Causality in political science

Darrell West. Air Wars (2013)Chattopadhyay and Duflo. Women as Policy Makers: Evidence from a Randomized Policy Experiment inIndia (2004)

Page 7: Causal inference in practice: Here, there, causality is everywhere

Causality in human behavior

Thistlewaithe and Campbell. Effect of public recognition of scholastic achievement (1960)

Christakis and Fowler. The collective dynamics of smoking in a large social network (2008)

Page 8: Causal inference in practice: Here, there, causality is everywhere

Causality in biology and medicine

Effect of Vitamin D deficiency on colon cancer

Effect of heart attack surgery on long-term health of patient

Page 9: Causal inference in practice: Here, there, causality is everywhere

Causality in web applications

Sharma and Cosley. Distinguishing between personal preference and homophily in online activity feeds (2016).

Sharma, Hofman and Watts. Estimating the causal impact of recommender systems (2015).

Page 10: Causal inference in practice: Here, there, causality is everywhere

Methods for estimating causal effects from observational data

Condition on observed covariates

• Stratification• Matching• Regression (?)

Mechanism-based strategies

• Path-based approaches

Natural experiments

• As-if experiments

• Instrumental Variables

• Regression discontinuity

Page 11: Causal inference in practice: Here, there, causality is everywhere

Towards unifying estimation strategies: Causal graphical models

Page 12: Causal inference in practice: Here, there, causality is everywhere

I. Ideal: Randomized experiments

Page 13: Causal inference in practice: Here, there, causality is everywhere

II. Conditioning on observed covariatesCorresponds to Backdoor criterion.

Page 14: Causal inference in practice: Here, there, causality is everywhere

a) StratificationCondition on different levels of socio-economic status.

Page 15: Causal inference in practice: Here, there, causality is everywhere

b) Matching Socio-Economic status is a function of parents’ income, locality and other observed indicators.

Page 16: Causal inference in practice: Here, there, causality is everywhere

b) Matching Model propensity to attend a particular school.

Pschool = f(PI, Loc, …)

Page 17: Causal inference in practice: Here, there, causality is everywhere

c) RegressionCondition on observed covariates by adding them as independent variables in regression.

Works only if true causalrelationship between variables is linear.

Page 18: Causal inference in practice: Here, there, causality is everywhere

III. Mechanism-based strategies Corresponds to Front door criterion.

Page 19: Causal inference in practice: Here, there, causality is everywhere

IV. Natural Experiments Look for experiments happening in the real world.

Promise greater generalizability than controlled lab experiments.

Require greater care to ensure validity of causal identification.

Page 20: Causal inference in practice: Here, there, causality is everywhere

a. (As-if) random experiments

Page 21: Causal inference in practice: Here, there, causality is everywhere

b) Instrumental variables

Shock! Increase in traffic

Page 22: Causal inference in practice: Here, there, causality is everywhere

c) Regression discontinuity

Page 23: Causal inference in practice: Here, there, causality is everywhere

The promise of graphical models

Which variables to condition on?

Page 24: Causal inference in practice: Here, there, causality is everywhere

Two graphical criteria explain all of conventional approaches A principled, succinct framework for causality.

Allows arbitrary functional forms for relationships between variables.

Leads to clear statements about causal assumptions.

If a causal effect can be identified, it can be derived using do-calculus (helpful for bigger graphs).

Page 25: Causal inference in practice: Here, there, causality is everywhere

Graphical models form a succinct, consistent and complete framework for causality.They are also practical.

thank you!Amit Sharmahttp://www.amitsharma.in