Welcome to the 5th course in our series on causal inference concepts and methods created by Duke University with support from eBay, Inc. Designed to teach you causal inference concepts, methods, and how to code in R with realistic data, this course focuses on how to use the advanced methods of instrumental variables and regression discontinuity to find causal effects. We will give you the reasoning be behind the methods, how you need to argue their validity, and what they look like in practice. We’ll stay away from dense statistical math and focus instead on higher level concepts that data scientists need to always consider when examining and making inferences about data. The course instructors and creators are Dr. Matt Masten (Duke University), James Speckart (Duke), Tyler Ransom (University of Oklahoma), Atilla Gyetvai (Duke) and Alexandra Cooper (Duke).
Introduction to Instrumental VariablesFree
This chapter will introduce you to Instrumental Variables (IV) analysis, a tricky but powerful method to find causality through indirect inferenceCourse Trailer for Causal Inference with R – Instrumental Variables & RDD50 xpThe Logic of Instrumental Variables50 xpVisual Logic of Instrumental Variables50 xpWhat is the Relationship Between the Instrument and the Outcome?50 xpWhat is the Relationship Between the Instrument and the Treatment?50 xpWhat is the Relationship Between the Instrument and the Other Pre-Treatment Variables?50 xpHow Do We Calculate a Causal Effect with Instrumental Variables Analysis?50 xpSome Instrumental Variables Terminology50 xpIs it Endogenous vs Exogenous Variation?50 xpInstrumental Variables in Action: Education and Wages (graphs)50 xpInstrumental Variables in Action: Education and Wages (tables)50 xpWhat if Our Data Shows No Clear Correlation Between Treatment and Outcome?50 xpThe Three Instrumental Variables Assumptions50 xpAlan the Egg Farmer Turns to Instrumental Variables50 xp
Instrumental Variables in PracticeFree
This chapter covers more details about Instrumental Variables Analysis, and allows you to practice instrumental variables for yourselfThe Problem of Weak Instruments50 xpAn Example of a Weak Instrument50 xpOverview of Research on Property Rights & Economic Development50 xpWhere Do Instruments Come From?50 xpUsing IV When There is Measurement Error50 xpLogical Arguments About Instrumental Variables50 xpLet's Code: Practice Using Instrumental Variables50 xpPractice Using Instrumental Variables: CreditCo100 xpRefutability and Nonrefutability of the Instrumental Variables Assumptions50 xpQuestioning the Validity of Instrumental Variables Results50 xpIndirect Inference50 xpLet's Code: Practice Computing Causal Effects Using Indirect Inference50 xpPractice Computing Causal Effects Using Indirect Inference: CreditCo100 xpTwo Stage Least Squares (2SLS)50 xpOverview of Squatters & Property Rights Experiment50 xpExample of IV: Property Rights & Market Beliefs50 xpLet's Code: Practice Computing Causal Effects Using 2SLS50 xpPractice Computing Causal Effects Using Two-Stage Least Squares (2SLS): CreditCo100 xpComparing the Estimates from Indirect Inference and 2SLS50 xpSolving Noncompliance in Experiments with Instrumental Variables50 xpOverview of the Oregon Healthcare Experiment50 xpUsing IV to Solve Noncompliance in the Oregon Heathcare Insurance Experiment50 xpBalancing Assumptions and Data When Using IV50 xpLet's Code: Practice Using IV to Solve Noncompliance in the OHIE50 xpPractice Using IV to Solve Noncompliance in the OHIE100 xp
Regression Discontinuity DesignFree
This chapter introduces you to Regression Discontinuity Design (RDD), which is often considered a more flexible relative of Instrumental VariablesRegression Discontinuity: Looking at People on the Edge50 xpRegression Discontinuity: More Analysis of Thistlethwaite and Campbell50 xpKey Variables in an RD Design50 xpIdentifying Key Variables in an RD Design50 xpHow to Compute Causal Effects in a Regression Discontinuity Analysis50 xpVisualizing RD Designs50 xpLet's Code: Examining Manipulation in Regression Discontinuity Design50 xpExamining Manipulation in Regression Discontinuity Design100 xpUsing RDD to Study Neighborhoods and Schools50 xpThinking about confounders in an RD design50 xpUsing Geographical Borders More Generally in RD Designs50 xpLet's Code: Practice Computing Regression Discontinuity Effects50 xpPractice Computing Regression Discontinuity Effects100 xp"Fuzzy" Regression Discontinuity: Addressing Blurry Lines Between Groups50 xpIdentifying Fuzzy and Sharp RDDs50 xpFuzzy RDD and Swiss Religion: Checking the Numbers50 xp
The Local Average Treatment Effect of IVFree
This chapter explains the treatment effect we are estimating with the instrumental variables method - LATE, or the Local Average Treatment EffectWhich Causal Effects Are We Actually Getting with Instrumental Variables?50 xpDefining LATE: The Local Average Treatment Effect50 xpLATE and the ATE for Always-Takers50 xpThe "No Defiers" Assumption50 xpComputing LATE, Part 1: Dividing the Population50 xpComputing LATE, Part 2: Finding the Compliers50 xpComputing LATE, Part 3: Getting a Result50 xpThe Pros & Cons of LATE50 xpInterpreting Different Results when Working with Noncompliance50 xpATEs, CATEs, and LATEs: What's the Difference?50 xpAn Alternate Way to LATE?50 xp
Matt MastenSee More
I'm an econometrician working on identification and causal inference.
SSRI @ Duke UniversitySee More
We create modular online educational content for Duke University's Social Science Research Institute to share with students and professionals around the world. If you have any feedback for us about any part of this course, please let us know!
Tyler RansomSee More
PhD in Economics
Alexandra CooperSee More
Alexandra Cooper earned a Ph.D. in political science from the University of North Carolina at Chapel Hill and has taught at Duke, UNC, UNC-Charlotte, and most recently, at Lafayette College, before joining Duke University to plan and develop the Social Science Research Institute (SSRI). Initially responsible for SSRI's day-to-day programming and operations, as the institute has grown and added staff, she has had the opportunity to focus on coordinating educational programming. Her key responsibilities include organizing workshops and events, and supporting programming focused on improving knowledge and applications of the tools and methods of the social and behavioral sciences.