Part V · Robustness & Production Deployment

Module 14. Partial Identification & Sensitivity Analysis

Replace fragile point estimates with honest ranges, and quantify how much hidden bias a finding survives.

Programme evaluation: high attrition in job training; robustness of observational findings about 6 hours3 lessons

Module progress0/3 lessons · 0%

Lessons

Partial Identification: Manski Bounds and Lee Bounds

· Explain partial identification: reporting the set of effects consistent with the data and weak assumptions
· Compute worst-case (Manski) bounds for a job-training programme with 30% attrition
· Tighten bounds with monotonicity assumptions, and see what each assumption buys

· Explain the sensitivity parameter Γ as a bound on hidden bias in treatment odds
· Build matched pairs from the LaLonde data
· Implement the Wilcoxon signed-rank Rosenbaum bound and find the Γ at which significance disappears

· Express sensitivity in terms of partial R² of an unobserved confounder (Cinelli and Hazlett)
· Compute the robustness value of a regression estimate and benchmark it against observed covariates
· Use coefficient stability (Oster's δ) to judge how much selection on unobservables would erase an effect