Intramed

Faculty of Medicine's internal website

Menu

Applied Epidemiology and statistics III: Causal inference with non-randomized data

Applied Epidemiology and statistics III: Causal inference with non-randomized data

3 credits (half-time)

Dates: November 2-27, 2020

Number of participants: 15

Location: Due to the ongoing corona pandemic, the course will be given online via Zoom

Schedule:

November 2-3: Individual preparations, such as learning basic Stata commands for those unfamiliar with this program

November 4-5: Lectures and exercises 13-17 on November 4, and at 9-17 on November 5

November 6-17: Self-studies and assignments

November 18-19: Lectures and exercises at 13-17 on November 18, and at 9-17 on November 19

November 20-27: Self-studies and take-home exam

Course teachers:

Ali Kiadaliri, associate professor, PhD, Department of Clinical Sciences, Lund University (ali.kiadaliri@med.lu.se)

Carl Bonander, associate professor, PhD, Department of Public Health and Community Medicine, University of Gothenburg (carl.bonander@gu.se)

Anton Nilsson, associate researcher, PhD, Department of Laboratory Medicine, Lund University (anton.nilsson@med.lu.se)

Course examiner:

Jonas Björk, professor, PhD, Department of Laboratory Medicine, Lund University (jonas.bjork@med.lu.se)

Language: English

Target group: Primarily PhD students in medicine but also post-docs, if places are available. Applicants should have passed Applied Statistics I and II, or equivalent.

Course content and aim: When using observational data rather than a randomized controlled experiment, it is often challenging to draw conclusions about cause and effect, because the treatment or exposure may be related to other factors that influence the outcome (i.e., confounding). This course presents several methods that may be used to produce credible estimates of causal effects also with observational data, particularly by exploiting random features in data, so-called quasi-experiments. The presented methods can be used in different applications, such as evaluations of treatments implemented in some geographic regions but not in others, treatments available to individuals that fall below/above some threshold (such as very low birth weight), or exposures that are genetically influenced (exploited in “Mendelian randomization”).

Specifically, the course has the following contents:

-  Causal inference according to the potential outcomes model

-  Intern and external validity, and threats against these

-  Adjustment for known confounders and the limitations of this approach

-  Overview of non-randomized (quasi-experimental) study designs, with an emphasis on

  • Instrumental variables analysis
  • Difference-in-differences design
  • Regression discontinuity design

Literature: A set of articles to be handed out during the course

Application form

Last day to apply is October 9.

Site overview