Estimating causal effects in randomised controlled trials with non-compliance

Specialeforsvar: Joao Pedro Marques Soares

Titel: Estimating causal effects in randomised controlled trials with non-compliance

Abstract: The presence of non-compliance in randomised controlled trials compromises the fair comparison between treatment groups achieved by randomisation, potentially resulting in biased estimates of the treatment effect. This thesis aims to compare standard approaches for estimating treatment effects -intention-to-treat, as-treated, and per-protocol - to the instrumental variables method, in the context of randomised controlled trials with non-compliance. We illustrate the methods with a simulation study as well as an application on data from a real trial. We find that the instrumental variables method,
although not perfect, can be a valuable addition to a standard analysis based on the intention-to-treat approach.

Vejleder: Anders Tolver
Censor:   Søren Andersen, NOVO