BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Optimization-based Sensitivity Analysis
DTSTART:20231215T100000
DTEND:20231215T110000
DTSTAMP:20260410T192824Z
UID:471bd8d90cd56a717416f75fd15c2fac497b01e4873f9f82151f05eb
CATEGORIES:Conferences - Seminars
DESCRIPTION:Tobias Freidling – Cambridge university\nPresentation in Ma
 thematics\nStatistics in general and causal inference in particular relies
  on untestable assumptions. Hence\, it is crucial to assess the robustness
  of obtained results to potential violations of these assumptions. However
 \, such sensitivity analysis is only occasionally undertaken in practice a
 s many existing methods only apply to relatively simple models and their r
 esults are often hard to interpret. This work takes a more flexible approa
 ch to sensitivity analysis by viewing it as a constrained stochastic optim
 ization problem.\nIn the talk\, I focus on the confounding influence of an
  unobserved covariate on estimating a causal effect of interest. In order 
 to assess deviations from the typical identification assumptions of ordina
 ry least squares (OLS) or two-stage least squares (TSLS) estimators\, the 
 causal effect is expressed in terms of partial correlations and standard d
 eviations. These can either be estimated from the observed data or serve a
 s sensitivity parameters. Practitioners can express their beliefs about th
 e values of these sensitivity parameters by stipulating constraints on the
 m. For instance\, they may compare the influence of the unmeasured confoun
 der on treatment or outcome with an observed variable in terms of R^2-valu
 es. The specified constraints render the causal effect partially identifie
 d. The range of feasible values can be estimated by solving a constrained 
 optimization problem minimizing or maximizing over the sensitivity paramet
 ers. In order to address uncertainty in the sensitivity analysis\, we cons
 truct confidence intervals for the partially identified range. Since a heu
 ristic "plug-in" confidence interval may not have any frequentist guarante
 es\, this work instead follows a bootstrap approach. This method will be i
 llustrated with a real data example and several user-friendly visualizatio
 n tools.\n\n 
LOCATION:https://epfl.zoom.us/j/67189839982
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR