Prerequisites: Introductory epidemiology. Introductory biostatistics or introduction to generalized linear models. Working knowledge of a general statistical software like SAS, Stata or R
Advisory Prerequisites: An introductory course on causal inference (e.g. EPID 780) is highly recommended
Description: This course introduces how to think about and conduct sensitivity analyses for uncontrolled confounding, selection bias and measurement error in epidemiologic studies. The course will demonstrate the intuition behind the separate and combined consequences of these sources of bias on estimating and inferring causal effects. It will provide practical quantitative skills for assessing the sensitivity of analytical results to these biases in order to aid credible causal modeling and inference using empirical epidemiologic studies
Learning Objectives: 1. Learn to articulate the different of impact of uncontrolled confounding, selection bias and measurement error separately and in combination.
2. Learn to depict visually these sources of bias and understand their impact using causal diagrams.
3. Learn to conduct quantitative bias analyses including multiple-bias modeling.
4. Learn to reason about and obtain bias parameters for sensitivity analyses.
5. Learn to apply and interpret probabilistic sensitivity analyses in epidemiologic studies.