Advanced methods for causal inference

MATH-655

Media

Media

This file is part of the content downloaded from Advanced methods for causal inference.
Course summary

Course Description

This year, we will consider causal inference in settings with interference. The course will be discussion-based and require that the student actively participate. Each week, the discussion will be organized around one paper, or sometimes a book chapter. Everybody is expected to have read the paper carefully before the session.

Evaluation 


To encourage participation and active learning, students will be required to actively present material in class and write one report, both will count towards the final degree.


Guideline for presenters

  • Start with why. Capture everyone’s interest by sharing why the paper is important to discuss. Before going into the details, give a summary of the key points of the paper. Include enough background information but avoid the urge to include every single detail at this point. However, we do aim to read the the technical results very closely; that is, you are expected to do a line by line review so that we really understand the results -- this is the main part of the presentation. 
  • Simplify complex information. You can, for example, create simple visual representations of difference ideas or techniques to help your audience understand the information. Avoid writing out complex information in heavy slides that nobody will read.
  • Give things space. Spend time and effort on the  important parts.
  • Present your opinions. Instead of simply summarizing, include your thoughts on all aspects of the paper to initiate a discussion. What were the strengths and weaknesses? What questions did you have when reading the paper? What is unclear? As the presenter, you are the “expert”. Share with the group the questions you came across yourself and any answers you may have found to address them. You can also ask the group for their thoughts to create a starting point for a conversation. Questions can be about methods, results, general ideas, impact and so forth. 


The course notes. 

Reading list (will be continuously updated)
  1. Hudgens, Michael G., and M. Elizabeth Halloran. "Toward causal inference with interference." Journal of the American Statistical Association 103.482 (2008): 832-842.
  2. Sävje, Fredrik, Peter Aronow, and Michael Hudgens. "Average treatment effects in the presence of unknown interference." Annals of statistics 49.2 (2021): 673.
  3. Hu, Yuchen, Shuangning Li, and Stefan Wager. "Average direct and indirect causal effects under interference." Biometrika 109.4 (2022): 1165-1172.
  4. Aronow, Peter M., and Cyrus Samii. "Estimating average causal effects under general interference, with application to a social network experiment." Annals of applied statistics (2017): 1912-1947.
  5. Ogburn, Elizabeth L., et al. "Causal inference for social network data." Journal of the American Statistical Association (2022): 1-15.
  6. Tchetgen Tchetgen, Eric J., Isabel R. Fulcher, and Ilya Shpitser. "Auto-g-computation of causal effects on a network." Journal of the American Statistical Association 116.534 (2021): 833-844.
  7. Lauritzen, Steffen L., and Thomas S. Richardson. "Chain graph models and their causal interpretations." Journal of the Royal Statistical Society Series B: Statistical Methodology 64.3 (2002): 321-348.
  8. Britton, Tom. "Stochastic epidemic models: a survey." Mathematical biosciences 225.1 (2010): 24-35.

No teaching. 

I will introduce the course and motivate our first article. 

The first article is of important historical interest, because it is considered to be one of the first (serious) articles on interference. Thus, this article has inspired many other contributions. Yet, you don't need to agree with everything the authors write!


We discuss the first article. 

We will finish the first article and then discuss the second article (Sävje et al., Annals of Statistics)

We continue discussing the details in the article by Sävje et al. Please read this paper carefully. I think there are many ideas and points to scrutinize. 

There are still many points to discuss in the article by Sävje et al.


When we are done, we continue with the article by Hu et al. (attached). 

Lorenzo will lead the discussion on the article by Sävje et al.

We will study the article by Aronow and Samii on Exposure Mappings. 

We will read the following article:

Ogburn, Elizabeth L., et al. "Causal inference for social network data." Journal of the American Statistical Association (2022): 1-15.

We will continue discussing the article by Ogburn, especially the estimation part. 

We will discuss Tchetgen Tchetgen, Eric J., Isabel R. Fulcher, and Ilya Shpitser. "Auto-g-computation of causal effects on a network." Journal of the American Statistical Association 116.534 (2021): 833-844.

We will read the following article on chain graphs: 

Lauritzen, Steffen L., and Thomas S. Richardson. "Chain graph models and their causal interpretations." Journal of the Royal Statistical Society Series B: Statistical Methodology 64.3 (2002): 321-348.

We will continue with the article on chain graphs. Furthermore, I have intentionally included a new article that you should read. This article does not concern interference in the classical statistical sense; it is a survey on standard methods to analyse infectious disease epidemics with stochastic models. The intention is to understand the different motivations / foundations for using such models. In particular, like some of the articles we have already studied, this week's article also aims to study effects of vaccination etc.  

Matias will present the paper on epidemic modelling