Wednesday 6th December 2017

John Paul Gosling - Associate Professor in Statistics at the University of Leeds

An introduction to Bayesian statistics

Bayesian methodologies are widely employed nowadays: sometimes for computational convenience and sometimes for philosophical reasons. In this talk, I will cover learning about the world in the light of evidence, the features of a Bayesian analysis, the various flavours of Bayesian methods and the computational methods employed. Throughout, I will attempt to contrast the Bayesian approach with other systems for inference and, hopefully, persuade you of Bayes’s natural role in learning and inference.

Duncan Wilson & Sam Hinsley - Leeds Institute of Clinical Trials

Application of a Bayesian approach to treatment selection in a rare disease sub-population

High risk multiple myeloma (HRMM) occurs in approximately 20-30% of MM patients and represents a rare sub-population with poor outcomes. A large national UK phase III trial, Myeloma XI/XI+, is currently evaluating treatment strategies for all patients with newly diagnosed MM. Diagnostic assessments have been undertaken in this trial to retrospectively identify high risk patients, but only limited data is available for the HRMM sub-population to date.

In this talk we will describe the development of a randomised phase II trial comparing two different approaches to treatment of HRMM. The trial presented a number of challenges in its design, including a rare patient population, variable standard treatment, and differing experimental treatment approaches requiring multiple endpoint evaluation. To address these challenges we adopted a Bayesian strategy using methods proposed by Thall, Simon and Estey (1995, 1996) and extended by Thall and Sung (1998) for randomised phase II selection trials. Treatment selection is based on monitoring multiple outcomes, with interim assessments incorporated to allow early termination of treatments for futility. Molecularly matched individual patient data from Myeloma XI/XI+ is used in place of a concurrent control arm, leading to a reduction in the required sample size.

We will present details of the study design, its implementation in practice and some extensions to the methodology available.

References:

Thall, PF, Simon, R, Estey, EH: Bayesian sequential monitoring designs for single-arm clinical trials with multiple outcomes. Statistics in Medicine 14:357-379, 1995

Thall, PF, Simon, R, Estey, EH: New statistical strategy for monitoring safety and efficacy in single-arm clinical trials. J. Clinical Oncology 14:296-303, 1996.

Thall, PF and Sung, H-G: Some extensions and applications of a Bayesian strategy for monitoring multiple outcomes in clinical trials. Statistics in Medicine 17:1563-1580, 1998.

Caitlin Buck - Professor in the School of Mathematics and Statistics at The University of Sheffield

Three decades of subjective Bayesian methods for radiocarbon dating

Chronology construction was one of the first applications used to show case the value of MCMC methods for Bayesian inference (Naylor and Smith, 1988; Buck et al, 1992). As a result of 25 years of on-going work by statisticians, software developers and users, Bayesian chronology construction is now ubiquitous in archaeology and is becoming increasingly popular in palaeoenvironmental research. In this talk I will argue (as did Steel, 2001) that this paradigm change resulted because the user communities concerned really are subjective Bayesians. I will showcase some of the existing applications and look five or so years into the future at the kinds of things that this user-community will need help with if they are to continue to be such a nice case study in the use of subjective Bayes.

References:

C.E. Buck, C.D. Litton, & A.F.M. Smith (1992) Calibration of radiocarbon results pertaining to related archaeological events, Journal of Archaeological Science, Vol. 19, Iss. 5, pp 497-512.

J. C. Naylor & A. F. M. Smith (1988) An Archaeological Inference Problem, Journal of the American Statistical Association, Vol. 83, Iss. 403, pp 588-595.

Steel, D. (2001). Bayesian statistics in radiocarbon calibration, Philosophy of Science, Vol 68, S153–S164.

Venue: Worsley 9.60 at the University of Leeds. Presentations 3-5pm.