Intention-to-Treat Analysis in Presence of Noncompliance

• Activity Number: 536
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Rajarshi Mukherjee*+
• Companies: Harvard University
• Keywords: Rubin's Causal Model ; Gibbs Sampling ; EM algorithm ; Bayesian p-value ; Non-Compliance

Abstract

The standard intention-to-treat (ITT) analysis focuses on the causal effect of assignment of treatment rather than the causal effect of actual receipt of treatment. This approach is valid for measuring the effect of the encouragement, but interest often centres on the biological effect of the treatment itself, and the interpretation of an ITT analysis is sometimes based on an implicit assumption that the effect of assignment is nearly the same as the effect of the treatment. So a more careful approach is needed to answer more specific causal inference questions. We have implemented Rubin's Causal Model for this purpose and demonstrated that the flu shot(influenza vaccination dataset reported by McDonals, C, Hiu, S,Tierney, W(1992)) have a positive effect on the health of the subjects in the study which involved complication of noncompliance.