Advanced Statistical Methods for Observational Studies

CHPR 290: Lecture Mondays 2:30-4:20pm. Class meets in

Mike Baiocchi MSOB x318, baiocchi{AT}stanford{DOT}edu Thursday 9 - 11 AM

David Rogosa Sequoia Hall 224, rag{AT}stanford{DOT}edu Monday 4:30 - 5:30

Course web page: http://rogosateaching.com/somgen290/

From explorecourses

CHPR 290: Advanced Statistical Methods for Observational Studies (EDUC 260B, STATS 266) Design principles and statistical methods for observational studies, particularly for cause and effect determinations. Topics include: matching methods, sensitivity analysis, instrumental variables, graphical models, marginal structural models. 3 unit registration requires a small project and presentation. Computing is in R. Pre-requisites: HRP 261 and 262 or STAT 209 ( HRP 239), or equivalent. See http://rogosateaching.com/somgen290/ Terms: Spr | Units: 2-3 | Grading: Medical Option (Med-Ltr-CR/NC)

Week 1 - Introduction, Potential Outcomes and Study Design. Week 2 - Causal Inference in Randomized Experiments and Models for Observational Studies Week 3 - Matching Methods and Implementations Week 4 - Further issues: Design Sensitivity Week 5 - Further issues: Planning Analysis Week 6 - Instrumental Variable Methods. a. encouragement designs and compliance adjustments b. fixing broken regressions: (IV illusion or promise)? Week 7 - Natural Experiments: Regression Discontinuity, Interrupted Time Series, Difference in Differences Week 8 - The Role of Graphical models, DAGs Week 9 - Applications of Marginal Structural Models.

In this course students will:

(1) Learn to identify key statistical issues in observational studies and methods and study designs to address issues of confounding.

(2) Become proficient with advanced statistical methods for observational studies: methods for missing data, matching based inference, sensitivity analysis, propensity score methods, instrumental variables, directed acyclic graphs, and marginal structural models. You should know which methods are useful in different situations, and which conditions have to be checked for the method to be applicable.

(3) Be able to perform detailed data analyses on a variety data using the statistical computation environment R. You should be able to implement all the methods presented in this course.

Prerequisites: HRP261 and 262, or Stat 209 (HRP 239) or by permission of instructor

Design of Observational Studies, Paul Rosenbaum, 1st Edition (Springer) Available online: Stanford access

Causal Inference, Miguel Hernan & Jamie Robins Available online: http://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/

Causal Inference in Statistics, Social and Biomedical Sciences: An Introduction, Guido Imbens and Don Rubin, 1st Edition (Cambridge University Press) Stanford access

We will make extensive use of the statistical computation and programming environment R. It is free and open source, and it has become the de facto standard for statistical analyses in most areas of academia and industry. Current version of R for spring quarter will be R version 3.2.4 to be released on 2016-03-10. For references and software: The R Project for Statistical Computing Closest download mirror is Berkeley The CRAN Task Views provide an organization and overview of the many R packages.

One specialized resource this course will draw from is Paul Rosenbaum's software page

Many students may also use or be interested in an editor for R (beyond emacs or command line). Professor Baiocchi will be using RStudio, download at .

Basic R References.

online Verzani book resources: version of text available from John Verzani's page . alternate version, single pdf UsingR R-package

The Registrar does have a form (no-fee) for faculty, staff, post-docs: Application for Auditor or Permit to Attend (PTA) Status