Sensitivity analysis for causal inference assumptions

Tools to assess the robustness of causal effect estimates to unmeasured confounding, model specification, and other threats to internal validity. Includes Rosenbaum bounds, the E-value family, Ding-VanderWeele bias formulas, tipping-point analysis, omitted- variable bias (Cinelli-Hazlett), Manski bounds, probabilistic (Monte-Carlo) bias analysis, and specification curve analysis.

Details

Wraps CRAN EValue, tipr, sensemakr, specr, rbounds, episensr, and konfound when available; falls back to base-R closed-form implementations otherwise.

References

Rosenbaum (2002); VanderWeele & Ding (2017); Cinelli & Hazlett (2020); Manski (1990); Ding & VanderWeele (2016).