BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Optimization-based Sensitivity Analysis for Unmeasured Confounding
  using Partial Correlations - Tobias Freidling (Statistical Laboratory)
DTSTART:20240126T153000Z
DTEND:20240126T170000Z
UID:TALK210868@talks.cam.ac.uk
CONTACT:97804
DESCRIPTION:Causal inference necessarily relies upon untestable assumption
 s\; hence\, it is crucial to assess the robustness of obtained results to 
 violations of identification assumptions. However\, such sensitivity analy
 sis is only occasionally undertaken in practice\, as many existing methods
  only apply to relatively simple models and their results are often diffic
 ult to interpret. We take a more flexible approach to sensitivity analysis
  and view it as a constrained stochastic optimization problem. This work f
 ocuses on sensitivity analysis for a linear causal effect when an unmeasur
 ed confounder and a potential instrument are present. We show how the bias
  of the OLS and TSLS estimands can be expressed in terms of partial correl
 ations. Leveraging the algebraic rules that relate different partial corre
 lations\, practitioners can specify intuitive sensitivity models which bou
 nd the bias. We further show that the heuristic ``plug-in'' sensitivity in
 terval may not have any confidence guarantees\; instead\, we propose a boo
 strap approach to construct sensitivity intervals which perform well in nu
 merical simulations. We illustrate the proposed methods with a real study 
 on the causal effect of education on earnings and provide user-friendly vi
 sualization tools. ("preprint":https://arxiv.org/abs/2301.00040)
LOCATION:Centre for Mathematical Sciences\, MR12
END:VEVENT
END:VCALENDAR