Machine Learning for Heterogenous Treatment Effect Estimation | Dr. Alicia Curth | TAP Seminar

Talk title:
Machine Learning for Heterogenous Treatment Effect Estimation: Understanding Challenges, Opportunities & Solutions

 
Abstract:
 
The need to evaluate treatment effectiveness is ubiquitous in most of empirical science, and interest in flexibly investigating effect heterogeneity from observational data is growing rapidly. Machine Learning methods, given their success on standard prediction tasks, are a natural candidate to become the basis of solutions for estimation of such heterogeneous treatment effects (HTEs). In this talk, we will first examine the unique machine learning challenges inherent to the HTE estimation problems – what sets HTE estimation apart from standard prediction problems? We will then focus on one challenge in particular – the absence of the true outcome label of interest due to the fundamental problem of causal inference – and discuss how this affects model design, model evaluation and model selection. Finally, we will briefly consider how the standard problem setup studied in the literature so far relates to the much broader class of HTE estimation problems that arise in health applications more generally.

Видео Machine Learning for Heterogenous Treatment Effect Estimation | Dr. Alicia Curth | TAP Seminar канала UoB BESTEAM Seminars