Propensity Score Methods for Causal Inference

Similar documents
Marginal, crude and conditional odds ratios

OUTCOME REGRESSION AND PROPENSITY SCORES (CHAPTER 15) BIOS Outcome regressions and propensity scores

Double Robustness. Bang and Robins (2005) Kang and Schafer (2007)

Estimating the Marginal Odds Ratio in Observational Studies

Propensity Score Weighting with Multilevel Data

Matching. Quiz 2. Matching. Quiz 2. Exact Matching. Estimand 2/25/14

Summary and discussion of The central role of the propensity score in observational studies for causal effects

An Introduction to Causal Analysis on Observational Data using Propensity Scores

Selection on Observables: Propensity Score Matching.

arxiv: v1 [stat.me] 15 May 2011

Marginal versus conditional effects: does it make a difference? Mireille Schnitzer, PhD Université de Montréal

Gov 2002: 5. Matching

Balancing Covariates via Propensity Score Weighting: The Overlap Weights

Propensity Score Analysis with Hierarchical Data

The Impact of Measurement Error on Propensity Score Analysis: An Empirical Investigation of Fallible Covariates

Since the seminal paper by Rosenbaum and Rubin (1983b) on propensity. Propensity Score Analysis. Concepts and Issues. Chapter 1. Wei Pan Haiyan Bai

Covariate Balancing Propensity Score for General Treatment Regimes

PROPENSITY SCORE MATCHING. Walter Leite

Causal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification. Todd MacKenzie, PhD

Balancing Covariates via Propensity Score Weighting

Causal Inference Basics

Causal Inference with a Continuous Treatment and Outcome: Alternative Estimators for Parametric Dose-Response Functions

Propensity Score for Causal Inference of Multiple and Multivalued Treatments

Weighting Methods. Harvard University STAT186/GOV2002 CAUSAL INFERENCE. Fall Kosuke Imai

Difference-in-Differences Methods

Propensity Score Methods, Models and Adjustment

Causal Inference with General Treatment Regimes: Generalizing the Propensity Score

Section 10: Inverse propensity score weighting (IPSW)

Empirical Likelihood Methods for Two-sample Problems with Data Missing-by-Design

Propensity Score Matching

Using Matching, Instrumental Variables and Control Functions to Estimate Economic Choice Models

Gov 2002: 13. Dynamic Causal Inference

Propensity Score Methods for Estimating Causal Effects from Complex Survey Data

Causal inference in multilevel data structures:

Propensity Score Matching and Analysis TEXAS EVALUATION NETWORK INSTITUTE AUSTIN, TX NOVEMBER 9, 2018

Studies. Frank E Harrell Jr. NIH AHRQ Methodologic Challenges in CER 2 December 2010

Rerandomization to Balance Covariates

Investigating mediation when counterfactuals are not metaphysical: Does sunlight exposure mediate the effect of eye-glasses on cataracts?

Quantitative Economics for the Evaluation of the European Policy

Weighting. Homework 2. Regression. Regression. Decisions Matching: Weighting (0) W i. (1) -å l i. )Y i. (1-W i 3/5/2014. (1) = Y i.

Gov 2002: 4. Observational Studies and Confounding

Vector-Based Kernel Weighting: A Simple Estimator for Improving Precision and Bias of Average Treatment Effects in Multiple Treatment Settings

Causal Inference Lecture Notes: Causal Inference with Repeated Measures in Observational Studies

What s New in Econometrics. Lecture 1

Empirical approaches in public economics

Bayesian causal forests: dealing with regularization induced confounding and shrinking towards homogeneous effects

MATCHING FOR EE AND DR IMPACTS

Research Design: Causal inference and counterfactuals

Strategy of Bayesian Propensity. Score Estimation Approach. in Observational Study

Front-Door Adjustment

Modeling Log Data from an Intelligent Tutor Experiment

Advanced Quantitative Research Methodology, Lecture Notes: Research Designs for Causal Inference 1

36-463/663: Multilevel & Hierarchical Models

Flexible Estimation of Treatment Effect Parameters

Primal-dual Covariate Balance and Minimal Double Robustness via Entropy Balancing

The problem of causality in microeconometrics.

Statistical Models for Causal Analysis

Combining Experimental and Non-Experimental Design in Causal Inference

Estimating and Using Propensity Score in Presence of Missing Background Data. An Application to Assess the Impact of Childbearing on Wellbeing

Econometrics with Observational Data. Introduction and Identification Todd Wagner February 1, 2017

Propensity Score Matching and Variations on the Balancing Test

CompSci Understanding Data: Theory and Applications

Propensity Score Matching and Variations on the Balancing Test

Dan Graham Professor of Statistical Modelling. Centre for Transport Studies

RECSM Working Paper Number 54 January 2018

Stratification and Weighting Via the Propensity Score in Estimation of Causal Treatment Effects: A Comparative Study

Ignoring the matching variables in cohort studies - when is it valid, and why?

e author and the promoter give permission to consult this master dissertation and to copy it or parts of it for personal use. Each other use falls

Authors and Affiliations: Nianbo Dong University of Missouri 14 Hill Hall, Columbia, MO Phone: (573)

Measuring Social Influence Without Bias

Rewrap ECON November 18, () Rewrap ECON 4135 November 18, / 35

NBER WORKING PAPER SERIES USE OF PROPENSITY SCORES IN NON-LINEAR RESPONSE MODELS: THE CASE FOR HEALTH CARE EXPENDITURES

RIETI Discussion Paper Series 15-E-090

Causal Inference in Observational Studies with Non-Binary Treatments. David A. van Dyk

University of California, Berkeley

WORKSHOP ON PRINCIPAL STRATIFICATION STANFORD UNIVERSITY, Luke W. Miratrix (Harvard University) Lindsay C. Page (University of Pittsburgh)

Logistic regression: Why we often can do what we think we can do. Maarten Buis 19 th UK Stata Users Group meeting, 10 Sept. 2015

The propensity score with continuous treatments

Estimating Causal Effects from Observational Data with the CAUSALTRT Procedure

Ratio of Mediator Probability Weighting for Estimating Natural Direct and Indirect Effects

Use of Matching Methods for Causal Inference in Experimental and Observational Studies. This Talk Draws on the Following Papers:

MMWS Software Program Manual

studies, situations (like an experiment) in which a group of units is exposed to a

Partially Identified Treatment Effects for Generalizability

Variable selection and machine learning methods in causal inference

Implementing Matching Estimators for. Average Treatment Effects in STATA

Use of Matching Methods for Causal Inference in Experimental and Observational Studies. This Talk Draws on the Following Papers:

Combining multiple observational data sources to estimate causal eects

Bootstrapping Sensitivity Analysis

Discussion of Papers on the Extensions of Propensity Score

An Empirical Comparison of Multiple Imputation Approaches for Treating Missing Data in Observational Studies

Estimating direct effects in cohort and case-control studies

Bounds on Causal Effects in Three-Arm Trials with Non-compliance. Jing Cheng Dylan Small

Data Integration for Big Data Analysis for finite population inference

Causal Analysis in Social Research

Biost 518 Applied Biostatistics II. Purpose of Statistics. First Stage of Scientific Investigation. Further Stages of Scientific Investigation

Targeted Maximum Likelihood Estimation in Safety Analysis

Geoffrey T. Wodtke. University of Toronto. Daniel Almirall. University of Michigan. Population Studies Center Research Report July 2015

Potential Outcomes and Causal Inference I

Behavioral Data Mining. Lecture 19 Regression and Causal Effects

Transcription:

John Pura BIOS790 October 2, 2015

Causal inference Philosophical problem, statistical solution Important in various disciplines (e.g. Koch s postulates, Bradford Hill criteria, Granger causality) Good reference on history of causal inference: Paul Holland Statistics and Causal Inference JASA, 1986

What can we estimate? Potential Outcomes Framework (Rubin s Causal Model) Notation: Z (1=treated, 0=control), baseline covariates X =(X 1,...,X p ), outcome Y potential outcomes Y 0, Y 1 We observe (Z, Y, X) foranindividual Y = ZY (1)+(1 Z)Y (0) Causal e ect of treatment: Y (1) Y (0) Average causal e ect: = E[Y (1) Y (0)]

What can we estimate? Average causal e ect ACE All or ATE = E[Y (1) Y (0)]) ACE Exp or ATT = E[Y (1) Y (0) Z = 1]) ACE Un or ATU = E[Y (1) Y (0) Z = 0]) Estimand and statistical methods depends on the study goal/question

Assumptions 1. ZprecedesY 2. Stable Unit Treatment Value Assumption (SUTVA) non-interference no variation in treatment 3. Strongly Ignorable Treatment Assigment (SITA) 0 < P(Z = 1 X) < 1 (this is the propensity score) (Y (0), Y (1)) Z X (very strong assumption) no unobserved confounders

Randomized Controlled Trials vs. Observational Studies RCTs Treatment e ects on outcome considered as causal Z is determined for each participant at random, (Y (0), Y (1)) Z E[(Y Z = 1) (Y Z = 0)] is unbiased estimate of ATT = ATE Observational Study Z is not controlled, (Y (0), Y (1)) Z E(Y Z = 1) =E(Y (1) Z = 1) = E(Y (1)). Cannot obtain unbiased estimate by direct comparison. But...

Potential Solution In observational studies, assuming SITA assumption is met then treatment assignment, Z, among individuals with particular X is essentially random and independent of potential outcomes Rosenbaum and Rubin (1983) - conditioning on the propensity score (PS) we can identify E(Y (0)) and E(Y (1)) from the observed data (Z, Y, X) andultimatelyestimate.

Propensity Score Austin, 2011: The propensity score is a balancing score: conditional on the propensity score, the distribution of observed baseline covariates will be similar between treated and untreated subjects This is a large sample property Unknown in practice, but can be estimated from the data, given some assumptions on e(x) (e.g. parametric regression model, CRTs). Mathematically: e(x) = P(Z = 1 X). R&Rshowedthat X Z e(x) and in addition to the SITA assumption, (Y (0), Y (1)) Z e(x). For theoretical properties see R&R (1983) and Lunceford and Davidian (2004)

The Propensity Score Model Goal: Covariate balance Popular method for estimating PS is logistic regression, though others exist (e.g. tree-based methods, random forests, neural networks, etc.) Regress logit[p(z = 1 X)] on X and obtain predicted probabilities (ê(x)) R&R (1984) and Austin 2011 describe an iterative approach: 1. Specify an initial model to estimate ê(x) 2. Perform diagnostics to assess covariate balance for each treatment 3. Modify PS by adding covariates, interactions, or using non-linear terms 4. Important: Each step should not be motivated by statistical significance but by objective

The Propensity Score Model Goal: Covariate balance What covariates do we include? Selection driven by subject-matter knowledge Only baseline variables Include all confounders and possible non-linear transformations (e.g. interactions). Overfitting generally not an issue (unless treatment is uncommon) Always include variables that a ect the outcome even if they don t a ect treatment assignment (Brookhart et al. (2006))

Diagnostics How do we know the PS model has been adequately specified? Assess standardized di erences of each covariate between treatment groups (very useful) Assess PS distributions by treatment (need common support condition) Compare distributions of the covariates between treatments Varies with PS method Di cult in practice with high dimensional data Assess the sensitivity of study conclusions to the SITA assumption.

Methods utilizing PS Matching Stratification Inverse PS weighting Covariate adjustment by PS PS methods allow for estimation of the marginal treatment e ect. The first three separate the design of the study from the analysis of the study. Can do subsequent regression adjustment to eliminate residual imbalance in prognostically important covariates after first three PS methods

Matching Simple formulation for ATT For each treated subject, select single untreated subject (without replacement) with same value of ê(x) or its logit (R&R, 1985) Take di erence of outcomes for the matched pair and average over all matched pairs Calculating ATE and ATU require slightly di erent sampling, possibly with replacement Advantage: Eliminates large proportion of systematic di erences in baseline characteristics between treated and untreated subjects Disadvantage: Inexact matching may lead to bias. Unmatched individuals are discarded, leading to loss in statistical power. Discarding individuals may also alter our estimand (Hill, 2008)

Stratification Easily estimate ATT: Create quantiles (e.g. quintiles) of the PS values, thereby dividing the subjects into equal-sized strata Within each stratum estimate treatment e ect Calculate weighted average of within-strata estimates of treatment e ect. Weight of each stratum is simply the percent of the quantile Estimating ATE and ATU require weighting by fraction of treated or untreated individuals, respectively, per stratum Advantage: Easy to construct and estimate causal e ects. Disadvantage: Small number of strata may result in residual confounding within the strata, resulting in bias. ATT estimates largely biased (compared to weighting)

Inverse weighting Weighted linear regression of outcome on treatment where w = Z w 1 + 1 Z w 2 For ATE, w 1 = e(x), w 2 = 1 e(x); foratt,w 1 = 1, w 2 = e(x) 1 e(x) ;ForATU,w 1 = 1 e(x) e(x), w 2 = 1. (Morgan & Todd, 2008) Advantage: Uses all available data; Can deal with more complex non-linear link functions (e.g. odds ratio); generally less biased than stratification (Lunceford & Davidian, 2004) Disadvantage: An individual with PS close to 0 or 1 will have unstable weights, leading to potentially spurious treatment e ects with high variance and wide CIs.

Covariate adjustment using PS Fit model: E(Y Z, X) = + Z + f (e(x)) (may include interaction of Z and e(x)) Can obtain ATE, ATT, and ATU by evaluating at di erent values of ê(x) Advantage: Allows for flexible relationship between PS and outcome (e.g. use of splines for PS) Disadvantage: Sensitive to whether PS has been accurately estimated. Analyst may be tempted to work toward desired or anticipated result, given that outcome is in sight.

Final Thoughts PS methods can be done without reference to outcome - i.e. separate study design from analysis Balance of covariates can be easily checked PS methods more robust to model misspecification compared to traditional outcome regression (all we care about is balance) Measures a di erent quantity, namely, the marginal/population treatment e ect (vs. conditional/individual treatment e ect in traditional regression) Important to distinguish the two in relation to study goals Omitted variable bias a ects internal validity of both approaches similarly Strategy so far is to balance covariates. Another idea is to find an instrument " S that is randomly assigned and a ects Y only through Z