Skip to content

lecture slide, tutorial code and dataset for 2023/06/30 Causal Machine Learning workshop

Notifications You must be signed in to change notification settings

Jinwoo-Yi/230630-GRF-workshop

Repository files navigation

Causal Machine Learning for Studying Individual Differences or Heterogeneous Treatment Effects

hosted by
Institute of Psychological Sciences, SNU
2023/06/30

led by
SNU Connectome Lab
/
PhD. Jiook Cha
Dept. Psychology
Seoul National University
/
BA. Jinwoo Yi
Dept. Brain & CogSci
Seoul National University

Abstract

Traditional causal inference methods focus on an average effect of a given factor or treatment on outcomes under the assumption of a constant causal effect in a given group. However, often important research questions arise from the individual differences of the causal or treatment effects. In many domains of science and modern data science , estimating this so-called heterogeneous treatment effects (or conditional average treatment effects) is key to data-driven decision making (e.g., treatment, intervention, or policy making) towards precision science. For example, identifying individuals with risk for suicides based on cognitive, behavioral, neural, genetic, environmental variables; recommending an effective treatment option for a patient based on the multi-modal data; predicting or stratifying an individual’s response to negative (e.g., traumatic) experiences. Fulfilling these tasks require more advanced analytics than mere mediation or moderation analysis.
In this 3-hour workshop, we will introduce the ‘generalized random forest (GRF) ’ as a data-driven computational learning approach to test individual differences or heterogeneity of a causal/treatment effect, and if so, to discover the set of variables contributing to the individual differences. GRF overcomes the limitations of conventional approaches that rely on predefined target variables to explain heterogeneity. Instead, GRF employs an ensemble of trees to maximize the differences in expected treatment effects, effectively identifying the key factors that contribute to detected heterogeneity. Participants will have opportunities to acquire key concepts of GRF and causal machine learning and to gain hands-on experience with GRF through tutorials. We will also share the examples regarding how GRF may be applied in behavioral and neural sciences. Participants will also have the chance to brainstorm and refine their own research questions that can be explored using GRF.

Timetable

[Part I, 14:00-15:10] LECTURE: Why Heterogeneity and Causal Machine Learning?
[Part II, 15:20-16:00] DISCUSSION: Which Questions can We Deal with GRF?
[Part III, 16:00-17:00] TUTORIAL: How to use GRF package?

Lecture Link

https://youtu.be/bCHVAjglvNE?si=6v4ri8HcTSJBOpsL

About

lecture slide, tutorial code and dataset for 2023/06/30 Causal Machine Learning workshop

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages