Software

RdistNN implements distributional nearest-neighbor methods for nonparametric regression, conditional average treatment effects (CATE), and conditional local average treatment effects (CLATE), connecting my research on local estimation and inference with applied workflows.

Software Packages

R

RdistNN for R

R package · Active development

An R package for local estimation with pointwise jackknife inference, tuning-sensitivity analysis, diagnostics, and ggplot2 visualizations.

  • Single- and two-scale nonparametric regression
  • Conditional average treatment effects (CATE)
  • Conditional local average treatment effects (CLATE)

Also includes simulation and affine-paper replication tools.

GitHub repository Documentation & examples
Stata

RdistNN for Stata

Stata companion · Stata 17+

The Stata companion provides all three workflows through native commands, with diagnostics, graphs, numeric exports, and prediction from saved fits.

  • Single- and two-scale nonparametric regression
  • Conditional average treatment effects (CATE)
  • Conditional local average treatment effects (CLATE)
GitHub repository Documentation & examples

Treatment-effect estimation uses supplied nuisance predictions; CLATE supports binary instruments and treatments. Installation instructions, examples, and statistical assumptions are covered in the documentation. PLQR and R-only ratio confidence sets remain experimental.

Related research: Jackknife variance estimation · Conditional inference without cross-fitting