The fortnightly BRAG meeting will be held this Thursday 2/11 at 1 pm via Zoom/GP-Y801. This week we will have presentations by @Navodya Abirami and @Adam.
Zoom Link: https://qut.zoom.us/j/7952714168
Abi’s Talk
Title: Robust methods to design Bayesian adaptive clinical trials.
Abstract: The outbreak of COVID-19 has brought about an unprecedented push to undertake clinical trial assessments as quickly as possible. Bayesian adaptive trials are particularly appealing in this regard as, based on the results of an interim analysis, without compromising scientific rigour or the validity of the trial. According to the results of an interim analysis, an adaptive trial: (1) allows early termination if a treatment demonstrates clear success or futility; (2) provides the flexibility to drop a treatment that is deemed ineffective; (3) can adapt to include new treatments as they become available; and (4) allows adjustments to treatment allocations such that treatments that appear more promising can be allocated more often, on average. However, many of the methods proposed for the adaptive design of clinical trials rely critically on assumptions about the data that may be observed during the trial. Unless these assumptions are correctly specified, they can lead to incorrect conclusions about the treatment effect. Thus, designing such trials poses challenges due to the inherent uncertainty surrounding the onset, course, and resolution of the disease, which motivates the needs to develop new approaches to design Bayesian adaptive trials that are robust or relax assumptions about the data that will be observed in the trial, and this is the overarching aim of this PhD project.
Adam’s Talk
Title: Transfer Sequential Monte Carlo: A Framework for Bayesian Model Transfer
Abstract: Model transfer attempts to incorporate information from related source domains to improve the inference on the target domain. Unfortunately, it is not clear when to transfer information, which information to transfer, and how to transfer this information. Current statistical model transfer methods are limited to conjugate distributions and suffer from some theoretical issues. We develop a new framework for Bayesian model transfer, transfer sequential Monte Carlo, that takes a principled statistical approach to the transfer problem. The new framework permits a convenient comparison of four different approaches for selecting the amount of tempering of source information, and does not require that the Bayesian model under consideration be conjugate. Two of these methods take a Bayesian model selection approach to the model transfer problem, by treating each level of transfer as a potential model to be selected. We explore two selection criteria, the model evidence and widely applicable information criterion. The second two methods consider the transfer factor to be a random variable and utilise the joint power prior and normalised power prior respectively. We use our new transfer sequential Monte Carlo framework to evaluate the efficacy of Bayesian model transfer, with the empirical results expanding on previously explored theoretical justifications and limitations.
Bayesian Research & Applications Group 