Course leader: Jonas Björk jonas.bjork@med.lu.se

Target Group
The course is intended for all doctoral students at the medical faculty that work with clinical or epidemiologic studies. The course is also open for postdoctoral researchers (course fee 5000 SEK, there is no fee for doctoral students)

Credits
3hp 

Language
English

Prerequisites
Applied Statistics I & Applied Statistics II (or corresponding knowledge obtained elsewhere)

Number of participants
20

Location
Six course days are planned in Lund 2018:

October 1 (full day) 
October 3 (full day) 
October 5 (half day) 
October 8 (full day) 
October 11 (full day) 
October 12 (half day).

Course content

This course focuses on propensity stratification and Mendelian randomization, two statistical approaches within causal inference that are widely used in clinical epidemiology to deal with measured and unmeasured confounding. The course includes three themes:

1) Introduction to causal inference

• Overview of different causal models and criteria for causality
• Counterfactual reasoning (potential outcomes)
• Causal diagrams (directed acyclic graphs)
• Commonly used statistical models

2) Propensity score methods

• Introduction to logistic regression
• Model specification
• Overview of methods for using propensity scores to analyze relationships
• Practical considerations. Pros and cons of the approach.

3) Mendelian randomization

• Introduction to instrumental variables. Examples of use.
• Overview of different models. Model specification.
• Practical considerations. Pros and cons of the approach

The course receives financial support from EpiHealth (Epidemiology for Health) and LUPOP (Lund University Population Research Platform).