Non-Parametric Bounds and Gibbs Sampler for Assessing Priming and Post-Treatment Bias

A set of tools to implement the non-parametric bounds and Bayesian methods for assessing post-treatment bias developed in Blackwell, Brown, Hill, Imai, and Yamamoto (2025) .

prepost

Overview

This package contains code and sample data to implement the non-parametric bounds and Bayesian methods for assessing priming and post-treatment bias in experimental studies under various assumptions.

To get started, please see the article that developed these methods:

• Blackwell, Matthew et al (2025). Priming Bias Versus Post-Treatment Bias in Experimental Designs. Political Analysis Published online 2025:1-17. doi:10.1017/pan.2025.3. (journal version, arXiv preprint)

Installation

## Install developer version
## install.packages("devtools")
devtools::install_github("mattblackwell/prepost", build_vignettes = TRUE)

Usage

Both the nonparametric and Bayesian estimators all have prefixes that indicate what type of experimental design being used.

• pre_ functions can analyze data from a pre-test design where the moderator is measured pre-treatment.
• post_ functions can analyze data from a post-test design where the moderator is measured post-treatment.
• prepost_ functions can analyze data from a random placement design, in which the moderator is randomly assigned to be measured before or after treatment.

Most functions can be specified with a formula to identify the outcome and treatment and another one-sided formula for the moderator:

library(prepost)
data(delponte)
out <- pre_bounds(
  formula = angry_bin ~ t_commonality,
   data = delponte,
  moderator = ~ itaid_bin
)
out

## $lower
##            
## -0.5923203 
## 
## $upper
##           
## 0.3221525 
## 
## $ci_lower
## [1] -0.6875343
## 
## $ci_upper
## [1] 0.4035053
## 
## $pre_est
## [1] -0.2701678