Estimate Bounds via Principal Stratification

When studying the causal effect of a treatment on an outcome, post-treatment events may render the outcome undefined. For example, the effect of job training on hourly wage is ill-defined for those who would not be employed without job training. This package estimates average causal effects among those whose outcome would be defined regardless of the intervention, producing a set-identified result. For more details, see Lundberg and Cho (2025) .

pstratreg

Regression-Based Principal Stratification

Ian updating as an example

WARNING. This website documents an R package that is in development. Expect changes. We encourage you to try the package and email us with suggestions. Project is joint work by Ian Lundberg ([email protected]) and Soonhong Cho ([email protected]).

This website documents the pstratreg package for R. The package helps study causal effects when some treatment values cause outcomes to be non-existent for some units.

Possible applications include

• a labor market intervention where the outcome is hourly wage, but some people are unemployed
• a medical intervention where the outcome is a health metric, but some people die before it is measured
• a sociological study where the outcome involves one's spouse, but some people divorce or never marry and thus have no spouse

This package provides regression-based methods for principal stratification that rely on parametric models and are useful in settings where one hopes to adjust for many confounders.

Getting started

To get started, first install R and RStudio. Then install the package.

devtools::install_github("ilundberg/pstratreg")

Now head on to the next page to see the types of questions this package can help answer.

This project is joint work between Ian Lundberg (Cornell, [email protected]) and Soonhong Cho (UCLA, [email protected]).