CAUSAL INFERENCE SEMINAR SERIES.

Instrumental Variables: Uncovering the Bounds of Possibility Joy Shi, Investigator, Mongan Institute, Massachusetts General Hospital Instrumental variables (IV) are powerful tools for estimating causal effects in the presence of unmeasured confounding, yet their full potential and limitations are seldom fully considered in applied settings. Using data from the NordICC trial of colonoscopy screening, this talk will discuss the breadth of IV methods, under a range of assumptions, for estimating both bounds and point estimates for the per-protocol effect of screening. I will emphasize how different IV assumptions shape our interpretations and the conclusions that can be drawn from the data. About the speaker Joy Shi is an Investigator at the Mongan Institute at Massachusetts General Hospital and CAUSALab at Harvard T.H. Chan School of Public Health. Her applied research leverages large administrative health databases to evaluate the comparative effectiveness of interventions for cancer and cardiovascular diseases, with recent work focusing on screening strategies for colorectal and prostate cancer. Methodologically, she specializes in the development of methods for instrumental variables and innovative applications of the target trial framework. Future seminars Ted Westling, University of Massachusetts Amherst, February 10, Charles River Campus CDS 365 Youjin Lee, Brown University, February 17, School of Public Health About the BU causal inference seminar The BU causal inference seminar series is co-organized by the Biostatistics Department (Sara Lodi and Fatema Shafie Khorassani) and the Mathematics & Statistics Department (Judith Lok). Lunch at 12:30 (email fshafie@bu.edu by 12/1 at noon to RSVP for lunch)