Advanced Approaches to the
Analysis of Observational Data
BIOSTAT 215 Spring 2012 (3 units)
Course Directors: Steve Shiboski, PhD
Professor, Department of Epidemiology & Biostatistics
Eric Vittinghoff, PhD
Professor, Department of Epidemiology & Biostatistics
Thomas B. Newman, MD, MPH
Professor of Epidemiology & Biostatistics and
Pediatrics
| OBJECTIVES |
 |
A common goal of observational clinical or epidemiologic research is to estimate the causal effect of particular exposures or interventions on some health outcome. While causation-oriented research has long been practiced, recent methodologic work has more sharply placed into view what it means and what is needed to estimate causal effects. In particular, it is now recognized that conventional stratification or regression approaches to reduce confounding in observational research may, in certain settings, fail to estimate unbiased causal effects. This course will describe more advanced methods that may succeed in estimating causal effects in cases where standard approaches break down.
At the end of the course, students will understand:
1. Potential outcomes, average causal effects, and local average causal effects;
2. Marginal and conditional causal effects;
3. Conditions where standard regression methods fail to estimate causal effects;
4. How and when to use propensity scores;
5. What to do when the time-dependent confounders of an intervention also mediate its effects;
6. How and when to use instrumental variables;
7. How to use treatment assignment as an instrument to estimate treatment efficacy; and
8. New-user analyses.
| PREREQUISITES |
|
Epidemiologic Methods
(Epi 203), Biostatistical
Methods II (Biostat 208), and Biostatistical
Methods III (Biostat 209). Exceptions to these prerequisites
may be made with the consent of the Course Director, space permitting. Contact Dr. Vittinghoff with questions.
| FACULTY |
|
| FORMAT |
|
Lectures: Wednesdays, 1:00 to 2:30 PM
Labs: Wednesdays, 2:30 to 4:00 PM
| TEXTBOOK |
|
The text book will be a draft chapter, which will be provided in .pdf form.
The statistical software package Stata (Stata Corporation, College Station,
Texas) is used in the program. The TICR Program has arranged for a sizeable discount for UCSF-affiliated personnel via the Stata GradPlan program.
| GRADING |
|
Grades will be based on 3 homework assignments.
| ENROLLMENT |
|
This course is open to a limited number of individuals outside
of the ATCR and Master's programs. Preference is given to UCSF-affiliated personnel.
We regret that auditing is not permitted. To apply for this course please fill
out and submit the application below. Cost and submission information are in
the application.
Application (Word doc, 20KB)
|
