rddsga

Weighted Subgroup Analysis in Regression Discontinuity Designs

Implements inverse probability weighted (IPW) subgroup analysis for regression discontinuity (RD) designs. When subgroups differ in observed moderators near the cutoff, a naive comparison of subgroup-specific RD estimates conflates the causal effect of the subgroup characteristic with the effect of correlated moderators. Reweighting observations via IPW balances observed moderators across subgroups at the cutoff, isolating the subgroup-attributable component of the treatment effect difference.

Both a Stata package and an R package are included in this repository.

---

Stata

Installation

From SSC:

ssc install rddsga

Latest version from this repository:

net from https://raw.githubusercontent.com/acarril/rddsga/main/stata/
net install rddsga
net get rddsga

Quick start

use rddsga_synth
rddsga Y Z X1 X2, sgroup(G) bwidth(10) reducedform bsreps(200)

See help rddsga for full documentation.

---

R

Installation

devtools::install_github("acarril/rddsga")

Quick start

library(rddsga)
data(rddsga_synth)

# Sharp RD with IPW balancing moderator m, bootstrap SEs
fit <- rddsga(
  y ~ m | sgroup,
  data    = rddsga_synth,
  running = ~ x,
  bwidth  = 0.5,
  bsreps  = 200,
  seed    = 42
)
print(fit)
summary(fit)   # also shows balance tables

See ?rddsga and vignette("rddsga-intro") for full documentation.

---

Reference

Carril, Alvaro, Andre Cazor, Maria Paula Gerardino, Stephan Litschig, and Dina Pomeranz. "Weighted Subgroup Analysis in Regression Discontinuity Designs." Working paper.

Issues

Please report bugs at https://github.com/acarril/rddsga/issues.