Causal Inference (CSI)

This unit covers modern statistical methods for assessing the causal effect of a treatment or exposure from randomised or observational studies.


Coordinator

Dr Jessica Kasza, Prof Andrew Forbes School of Public Health and Preventive Medicine, Monash University


COORDINATORS:
Dr Jessica Kasza Monash University, Department of Epidemiology and Preventive Medicine Semester 2
Prof Andrew Forbes Monash University, Department of Epidemiology and Preventive Medicine Semester 2
General outline

Prerequisites

Epidemiology, Mathematical Background for Biostatistics, Probability and Distribution Theory, and Linear Models or a multivariable regression unit of study from a Master of Public Health course or equivalent

Time commitment

8-12 hours total study time per week

Semester availability

Semester 2

Assessment

Two major assignments worth 30% each, and 4 shorter assignments worth 10% each concerning concepts, derivations or applications.

Prescribed Texts

Hernán MA, Robins JM (2018). Causal Inference. Boca Raton: Chapman & Hall/CRC, forthcoming [free to download (as of April 2020) https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/ ] For details, including ISBN, see the BCA Textbook and Software Guide

Special Computer Requirements

Stata and R statistical software

Content

The unit begins by explaining the fundamental concept of counterfactual or potential outcomes and introduces causal diagrams (or directed acyclic graphs (DAGs)) to visually identify confounding, selection and other biases that prevent unbiased estimation of causal effects. Key issues in defining causal effects that are able to be estimated in a range of contexts are presented using the concept of the “target trial” to clarify exactly what the analysis seeks to estimate. A range of statistical methods for analysing data to produce estimates of causal effects are then introduced. Propensity score and related methods for estimating the causal effect of a single time point exposure are presented, together with extensions to longitudinal data with multiple exposure measurements, and methods to assess whether the effect of an exposure on an outcome is mediated by one or more intermediate variables. Comparisons will be made throughout with “conventional” statistical methods. Emphasis will be placed on interpretation of results and understanding the assumptions required to allow causal conclusions. Stata and R software will be used to apply the methods to real study datasets.

Special Computer Requirements

Course notes, assignment material and interaction facilities available online