Registration for a live webinar on 'Chronic inflammation, immune cell trafficking and anti-trafficking agents' is now open.See webinar details
Bayesian methods in health economics: prior distributions 2
A selection of talks on Methods
Artificial intelligence in medicine: history & state of the art
- Prof. John Fox
- University of Oxford, UK
Understanding statistics in epidemics and pandemics: lessons learned from COVID-19
- Prof. Sarah Ransdell
- Nova Southeastern University, USA
International biobanking: overview of key practices and policies
- Dr. Jim Vaught
- Editor in Chief, Biopreservation & Biobanking, USA
Hello, I'm Tony O'Hagan. Welcome to the second talk in my series, Bayesian Methods in Health Economics. This talk is entitled Prior Distributions.
Let's recap on the outline of the course as a whole. The first talk was on Bayesian principles, and from that you should have learned about the basic ideas of the Bayesian approach to statistics, and how it differs particularly from the frequentist approach which is more familiar to most people. In this second talk, we'll deal with prior distributions, which are something unique that goes into a Bayesian analysis. It's very important to think about those properly. These first two talks make up the section that's about Bayesian methods in general. Then we move on to the particulars of how we use Bayesian methods in health economic evaluation. The first of those, talk three, is about uncertainty in health economic evaluation. Then we move on to dealing with specifically probabilistic sensitivity analysis, which is a way of coping with and quantifying those uncertainties. Finally, the fifth talk is about formulating input uncertainties to go into that analysis. Those last three parts as a whole deal with the whole ideas of Bayesian methods in health economic evaluation.
Here is the plan of this talk. We will begin by briefly reviewing the idea of prior information, and then move on to the topic of weak prior distributions. Here, the idea is to represent the fact that we have minimal prior information relative to what we're going to get from the data. Having dealt with that case, we move on to the contrary situation where we do have useful prior information that will contribute to the analysis, and we want to formulate informative priors to represent it. The key way of doing that is through the process called elicitation. Then there'll be a case study of using elicitation. Which leads us on to the situation where we have many parameters and we need to structure the prior information appropriately. That's the idea of hierarchical modelling. Prior information comprises everything