causalfrag

A Cross-Framework Causal Fragility Index for Sensitivity Analysis in Observational Studies

Overview

causalfrag provides a unified workflow for running, classifying, visualizing, and interpreting sensitivity analyses for unmeasured confounding across multiple causal frameworks.

The core contribution is the Causal Fragility Index (CFI): a single 0--100 composite score that integrates evidence from multiple sensitivity frameworks into one interpretable measure of robustness.

Relationship to confoundvis

confoundvis (Hait, 2026) provides visualization tools for sensitivity analysis. causalfrag provides the unified fragility scoring, interpretation, and reporting layer that optionally uses confoundvis graphics.

Installation

# CRAN (once available)
install.packages("causalfrag")

# Development version
remotes::install_github("causalfragility-lab/causalfrag")

Quick start

library(causalfrag)

fit <- lm(peacefactor ~ directlyharmed + age + female, data = darfur)

res <- run_sensitivity(fit, treatment = "directlyharmed", data = darfur)
res <- flag_fragility(res)   # CFI computed automatically

print(res)       # full output with CFI score
print_cfi(res)   # dedicated CFI display
generate_report(res)  # full structured report

Causal Fragility Index

Component	Framework	Full score at
RV (point est.)	sensemakr	RV >= 0.20
RV (significance)	sensemakr	RV >= 0.10
E-value	EValue	E-value >= 3.0
Pct bias	konfound	pct >= 100%

CFI range	Label
0 -- 24	Fragile
25 -- 49	Moderately fragile
50 -- 74	Moderately robust
75 -- 100	Robust

Author

Subir Hait -- Michigan State University
haitsubi@msu.edu
GitHub: causalfragility-lab