Methodology of causal inference when an unmeasured confounder exists

2014 Fiscal Year Final Research Report

• Project/Area Number: 23700344
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: Kinki University
• Principal Investigator: CHIBA Yasutaka 近畿大学, 医学部附属病院, 講師 (80362474)
• Project Period (FY): 2011-04-28 – 2015-03-31
• Keywords: 因果推論

Outline of Final Research Achievements

I proposed some sensitivity analysis methods for unmeasured confounders. In addition, I discussed a relationship between two conditions that can be used to derive the bounds on which the casual effect must be included, and compared these two conditions. When an instrumental variable exists, I derived the bounds for causal effects with several populations as the standard population, by applying the concept of principal stratification.
I presented the bounds for the principal strata effect when an unmeasured confounder exists. Furthermore, I showed that principal strata effects correspond to direct and indirect effects under special assumptions including the monotonicity assumption. I derived the bound for direct and indirect effects the sensitivity analysis methods. For the derivation, I applied those for the principal strata effect.

Free Research Field

統計科学