Old Dog, New Tricks: A Principled Approach to Machine Learning for Experiments

• In Event: Experimental Design and Analysis

Brief Overview

We outline a wide range of machine learning applications for experiments including: detecting heterogeneity in treatment effects; assessing randomization and balance; detecting and addressing subject attrition; and using unsupervised scaling methods.

Authors

• Jack Rametta, University of California-Davis

• Sam Fuller, University of California-Davis