[Case Study] Treatment Effect Heterogeneity Bias

[PhilippBach]

1. Description of the Case Study

A simulation-based illustration of the treatment effect heterogeneity bias

2. Idea for the data product

The data product could be an app, report or blog post. It could contain the following components

Simulation - Data generating process

• A DGP with potential outcomes for a binary treatment varible ($Y^{a=1}, Y^{a=0}$)
• Two (or more) scenarios with and without treatment heterogeneity

Presentation of results

• Introduction and explanation, maybe based on decomposition (see Section 4.3.1 of Cunningham (2020) or Section 2 in Huber (2021)
• An illustration (boxplot, density plot/histogram, etc.) of treatment effect heterogeneity
• Illustration of heterogeneity bias (regression output, regression line & scatter plot, etc)

3. Available resources and references

• Section 2 of Huber (2021)
• Section 4.3.1 of Cunningham (2020)

4. Comments

• The focus of this study is probably to find a nice simulation example illustrating the treatment effect heterogeneity bias and illustrate the consequences with tools in R (plots etc.)