CAUSAL INFERENCE IN MEDICINE AND PUBLIC HEALTH II Syllabus
Course Learning Objectives
Upon successfully completing this course, students will be able to:
- describe causal problems as potential interventions, through the framework of potential outcomes and assignment mechanisms
- discuss the role of designs and of different modes of statistical inference
- implement efficient (likelihood) methods with ignorable assignment of treatments,
- describe the role of outcome models and of propensity score models
- Assess when and how comparisons of longitudinal treatments can be designed as having sequentially ignorable assignment, and learn ways to estimate their causal effects
- Master efficient methods for estimating effects in studies with noncompliance to treatment, direct and indirect effects, and censoring by death
Course DescriptionPresents principles, methods, and applications in drawing cause-effect inferences with a focus on the health sciences. Building on the basis of 140.664, emphasizes statistical theory and design and addresses complications and extensions, aiming at cultivating students’ research skills in this area. Includes: detailed role of design for causal inference; role of models and likelihood perspective for ignorable treatment assignment; estimation of noncollapsible causal effects; statistical theory of propensity scores; use of propensity scores for estimating effect modification and for comparing multiple treatments while addressing regression to the mean; theory and methods of evaluating longitudinal treatments, including the role of sequentially ignorable designs and propensity scores; likelihood theory for instrumental variables and principal stratification designs and methods to deal with treatment noncompliance, direct and indirect effects, and censoring by death.
Methods of AssessmentStudent evaluation based on problem sets and a final project.
Prerequisites140.654 or equivalent for matrix representation of multiple linear and logistic regression
Please see the course Session for a full list of dates and items for this course.
Academic Ethics Code
Students enrolled in the Bloomberg School of Public Health of The Johns Hopkins University assume an obligation to conduct themselves in a manner appropriate to the University's mission as an institution of higher education. A student is obligated to refrain from acts which he or she knows, or under the circumstances has reason to know, impair the academic integrity of the University. Violations of academic integrity include, but are not limited to: cheating; plagiarism; knowingly furnishing false information to any agent of the University for inclusion in the academic record; violation of the rights and welfare of animal or human subjects in research; and misconduct as a member of either School or University committees or recognized groups or organizations.
Syllabus for 140.665: Experimental and non-experimental designs for
estimating causal effects
Part 2 of 2-course sequence
Lectures: Tuesdays, Thursdays: 15:30 – 16:50, W2009
Lab: Wednesdays of April 13, 27, and May 4, 9:30-10:30, W3017
Instructor: Constantine Frangakis, email@example.com.Office Hours: Fridays 1-2PM, E3642
Michael Rosenblum, firstname.lastname@example.org, Office Hours: Thursdays 9-10AM, E3616
Prerequisites:140.651-654 and 140.664, or consent of the instructor
Evaluation: three problem sets and a project
The instructors acknowledge sharing of valuable ideas and material with Donald Rubin and Guido Imbens.
Presents principles, methods, and applications in drawing cause-effect inferences with a focus on the health sciences. Building on the basis of Part 1 of the sequence, Part 2 emphasizes statistical theory and design and addresses complications and extensions, aiming at cultivating students’ research skills in this area. Includes: (1) examples showing the critical role of the design for causal inference; (2) role of models and likelihood perspective for ignorable treatment assignment; (3) importance and methods of using parametric models while preserving nonparametric validity; (4) estimation of noncollapsible causal effects; (5) statistical theory of propensity scores; use of propensity scores for estimating effect modification and for comparing multiple treatments while addressing regression to the mean; (6) importance and methods of combining propensity scores and outcome models; (7) theory and methods of evaluating longitudinal treatments, including the role of sequentially ignorable designs and propensity scores; (8) likelihood theory for instrumental variables; and (9) principal stratification designs and methods to deal with treatment noncompliance, vaccine effects, and mendelian randomization.
Course learning objectives:
Upon successful completion of this course, students will 1) Understand causal problems as potential interventions, through the framework of potential outcomes and assignment mechanisms, 2) Understand the role of designs and of different modes of statistical inference, 3) Understand and be able to implement efficient (likelihood) methods with ignorable assignment of treatments, 4) Understand the role of outcome models and of propensity score models, 5) Understand when and how comparisons of longitudinal treatments can be designed as having sequentially ignorable assignment, and learn ways to estimate their causal effects, and 6) Master efficient methods for estimating effects in studies with noncompliance to treatment, direct and indirect effects; censoring by death; and mendelian randomization.
Assignment Due Dates:
April 7 Problem Set 1 Due
April 21 Problem Set 2 Due
May 5 Problem Set 3 Due
May 20 Project Due
Readings: Available on the Courseplus Website
Topics covered:given below,where the context reviewed from part 1 (Elizabeth Stuart’s 3rd quarter course) is shown in normal italics font, and the topics emphasized in this part are in bold font
Chapter 1. Introduction and Framework – the critical role of the design
(1) Definition of potential outcomes, causal effects, the assignment mechanism as structuring concept;
(2) Example of the critical role of using an appropriate design-assignment mechanism: the case-crossover design and its analysis
(3) Outline of course
*Holland, P. (1986). Statistics and causal inference, (with discussion and rejoinder), Journal of theAmerican Statistical Association, 81: 945-970.
*Neyman (1923, pp. 465-468). On the application of probability theory to agricultural experiments. Essay on Principles. Section 9, translated in Statistical Science (with discussion) 5(4): 465-480, 1990.
*Rubin, D.B. (1990). Comment: Neyman (1923) and causal inference in experiments and observational studies. Statistical Science 5(4): 472-480.
*Rubin, D.B. (1974). Estimating causal effects of treatments in randomized and non-randomized studies. Journal of Educational Psychology 66: 688-701.
Varadhan, R, and Frangakis, CE (2004). Revealing and addressing length-bias and heterogeneous effects in frequency case-crossover studies. American Journal of Epidemiology 159, 596–602.
Chapter 2.Randomized experiments – use of parametric models with robust validity
(1) Fisher’s mode of inference focusing on the null hypothesis
Using parametric models with non-parametric validity – the case with randomization inference
(2) Neyman’s mode of inference
(3) Likelihood mode of inference for large population
(4) Using parametric models with non-parametric validity – the case with large population inference:
4.1 a class of GLM models with robust inference
4.2 a broad symmetry characterization of models with robust inference
(5) heterogeneous probabilities of assignment and the Horvitz-Thompson estimator
Fisher, R. A. (1925). Statistical Methods for Research Workers, 1st ed. Edinburgh: Oliver and Boyd.
Fisher, R. A. (1947). The Design of Experiments, 4th ed. New York: Hafner-Publishing.
*Neyman, J. (1923). On the Application of Probability Theory to Agricultural Experiments. Essay on Principles. Section 9, translated in Statistical Science, (with discussion), Vol 5, No 4, 465--480, 1990. (1923, pp. 468-472).
*Petersen, M. Wang, Y, van der Laan, M.J.Assessing the Effectiveness of Antiretroviral
Adherence Interventions Using Marginal Structural Models to Replicate the Findings ofRandomized Controlled Trials. (J Acquir Immune Defic Syndr 2006;43:S96–S103)
*Rubin, D.B. (1978). Bayesian inference for causal effects: The role of randomization. The Annals of Statistics 6: 34-58.
Rosenbaum P. R. (2002). Covariance adjustment in randomized experiments and observational studies. Statistical Science 17:286-327.
*Rosenblum M, van der Laan MJ. Using Regression Models to Analyze Randomized Trials: Asymptotically Valid Hypothesis Tests Despite Incorrectly Speciﬁed Models.Biometrics. 2009; 65(3): 937-945
*Rosenblum, M, and van der Laan, M.J. (2010) "Simple, Efficient Estimators of Treatment Effects in Randomized Trials Using Generalized Linear Models to Leverage Baseline Variables," The International Journal of Biostatistics: Vol. 6: Iss. 1, Article 13. Available at: http://www.bepress.com/ijb/vol6/iss1/13
Chapter 3. Ignorable treatment assignment and motivation for propensity scores
(1) Ignorable designs and likelihood methods
(2) Estimating noncollapsible population effects
(3) Small numbers of covariates; complications with many covariates; propensity scores.
Imbens, G.W. (2004). Nonparametric estimation of average treatment effects under exogeneity : A review. Review of Economics and Statistics 86(1) : 4-29.
Chapter 4. Propensity scores : theory and implementation
(1) Propensity scores: theory and diagnostics
(2) Use of propensity scores for estimating effect modification
(3) Propensity scores for comparing multiple treatments while addressing regression to the mean
(4) The advantages of combining use of propensity scores and models for the outcomes
*Rosenbaum, P.R. and Rubin, D.B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika 70(1) : 41-55.
*Rosenbaum, P., and Rubin, D. B. (1984). Reducing Bias in Observational Studies Using Subclassification on the Propensity Score,'' Journal of the American Statistical Association, Vol 79, 516--524.
*Lalonde, R. (1986). Evaluating the Econometric Evaluations of Training Programs. American Economic Review, Vol 76, 4, 605--620.
*Dehejia, H.R. and Wahba, S. (1999). Causal effects in non-experimental studies: Reevaluating the evaluation of training programs. Journal of the American Statistical Association 94: 1053-1062.
*Imbens, G.W. (2000). The role of the propensity score in estimating dose-response functions. Biometrika 87: 706-710.
Imai, K. and van Dyk, D.A. (2004). Causal inference with general treatment regimes: Generalizing the propensity score. Journal of the American Statistical Association 99(467): 854-866.
Huang, I.C., Frangakis, C.E., Dominici, F, Diette, G.B. and Wu, A.W. (2005). Application of a propensity score approach for risk adjustment in profiling multiple physician groups on asthma care. Health Serv Res 40(1): 253-278.
*Joseph L. Schafer and Joseph Kang (2008). Average Causal Effects From Nonrandomized Studies: A Practical Guide and Simulated Example. Psychological Methods. Vol. 13, No. 4, 279–313
Chapter 5. Methods and implementation of sensitivity analysis
Assessing sensitivity to the assumption of ignorable treatment assignment.
*Rosenbaum, P.R. and Rubin, D.B. (1983). Assessing sensitivity to an unobserved binary covariate in an observational study with binary outcome. Journal of the Royal Statistical Society, Series B, 45: 212-218.
Rosenbaum P. R. (2002). Covariance adjustment in randomized experiments and observational studies. Statistical Science 17:286-327.
Chapter 6. Studies with longitudinal treatments: theory and methods
(1) Examples of studies and goals: three case-studies
(2) Sequential ignorability as a formulation of having “access to physicians’ rules” for changing treatment recommendations;
(3) Estimability of estimands with few time points
(4) Inferential methods: conditional models, longitudinal propensity scores; marginal structural models; structural nested models
*Robins, J.M. (1987). A graphical approach to the identification and estimation of causal parameters in mortality studies with sustained exposure periods. Journal of Chronic Diseases 40: 1395-1615.
Robins JM, Greenland S. (1994). Adjusting for differential rates of PCP prophylaxis in high-versus low-dose AZT treatment arms in an AIDS randomized trial. Journal of the American Statistical Association, 89: 737-749.
Greenland, S., Pearl, J., and Robins, J.M. (1999). Causal diagrams for empirical
Research. Epidemiology 10: 37-48.
*Robins, J.N., Hernan, M.A. and Brumback, B. (2000). Marginal structural models and causal inference in epidemiology. Epidemiology 11: 550-560.
*Achy-Brou, AC, Frangakis, CE, and Griswold, M. Evaluating longitudinal treatments using regression models on propensity scores. Biometrics in press.
*Orellana, L., Rotnitzky, A, and Robins, J.M. (2010) "Dynamic Regime Marginal Structural Mean Models for Estimation of Optimal Dynamic Treatment Regimes, Part I: Main Content," The International Journal of Biostatistics: Vol. 6: Iss. 2, Article 8. Available at: http://www.bepress.com/ijb/vol6/iss2/8
*Cain, L.E., Robins, J.M., Lanoy, E., Logan, R., Costagliola, D., and Hernán, M.A. (2010) "When to Start Treatment? A Systematic Approach to the Comparison of Dynamic Regimes Using Observational Data," The International Journal of Biostatistics: Vol. 6: Iss. 2, Article 18. Available at: http://www.bepress.com/ijb/vol6/iss2/18
Chapter 7. Studies with nonignorable noncompliance and instrumental variables
(1) Review of assumptions of instrumental variables from part 1.
(2) Non-linear estimands; likelihood estimation and the EM algorithm
(3) Studies with non-compliance to treatment and incomplete outcomes
*Angrist, J., Imbens, G.W., and Rubin, D.B. (1996). Identification of causal effects using instrumental variables (with discussion). Journal of the American Statistical Association 91(434): 444-472.
*Imbens, G., and Rubin, D. B. (1997). Bayesian Inference for Causal Effects in Randomized Experiments with Noncompliance. The Annals of Statistics, Vol 25, 305--327.
*Sommer, A. and Zeger, S. L. (1991). On Estimating Efficacy from Clinical Trials. Statistics in Medicine, 10, 45--52.
*Frangakis, C.E., Brookmeyer, R.S., Varadhan, R., Mahboobeh, S., Vlahov, D., and Strathdee, S.A. (2004). Methodology for evaluating a partially controlled longitudinal treatment using principal stratification, with application to a Needle Exchange Program. Journal of the American Statistical Association 99: 239-249.
Chapter 8. More general partially controlled studies and principal stratification
(1) Partially controlled studies as a general class of problems;
(2) Principal stratification as a framework for formulating questions in partially controlled studies
(3) Censoring by death
(4) Vaccine studies and effect of cell-mediated immunity vaccines
(5) Case-control and other designs with instrumental variables
(6) Mendelian randomization
*Frangakis, C.E. and Rubin, D.B. (2002). Principal stratification in causal inference. Biometrics 58: 21-29.
*Zhang, J.L., and Rubin, D.B. (2003). Estimation of causal effects via principal
stratification when some outcomes are truncated by ‘death.’ Journal of Educational and Behavioral Statistics 28: 353-368.
*Gilbert, P.B., Bosch, R.J., and Hudgens, M.G. (2003). Sensitivity analysis for the assessment of causal vaccine effects on viral load in AIDS vaccine trials. Biometrics 59: 531-541.
Disability Support ServicesIf you are a student with a documented disability who requires an academic accommodation, please contact the Office of Student Life Services at 410-955-3034 or via email at email@example.com.