Registration for a live webinar on 'Chronic inflammation, immune cell trafficking and anti-trafficking agents' is now open.See webinar details
Bayesian adaptive designs for clinical trials
Published on December 29, 2016 42 min
Other Talks in the Series: Adaptive Clinical Trial Design
Frequentist approaches: sample size in adaptive clinical designs
- Prof. Tatsuki Koyama
- Vanderbilt University, USA
My name is Ben Saville. I'm a statistical scientist from Berry Consultants. I'm also an adjunct assistant professor at the Vanderbilt University School of Medicine in Nashville, Tennessee. And this is a talk on Bayesian Adaptive Designs for Clinical Trials.
Many of us have probably moved to a new place or a new city and started a new job, and you might be posed with the question of how do you get to work and you know there are different options, of course.
You could take a highway, you could take the back roads, you have to figure out which way you're going to go. You want to go the optimal way to get you to work as fast as you can and without stressing you out too much. There are different solutions to decide how to do this. One possible solution which I'm sure we've all done is you take 30 envelopes in your car with you and it has one of three routes on it and you randomly select an envelope and you drive that route. Record the drive time and at the end of the 30 days you look at your data, that point you drive whichever route had the fastest average drive time. I'm sure we've all done that and I'm being facetious here because obviously that's something that probably no one would do. The more intuitive way that mostly all of us would decide how to drive to work would be to pick one route, you drive it the first day, you make a note of how long it took you and then maybe the next day you drive a different way. And if it took you longer than the first day then maybe you switch back to the first and we iterate through this process of trying different ways.
Really, we tend to do more of an adaptive approach to figuring out which way to drive to work. If one route takes too long, we're probably less likely to try it again and if we take it twice and it takes a long time both times, then we're probably going to drop it. If you drive a certain way one way and it takes you forever, you may never even go that way again. So basically, the idea here is that there's a point at which you feel like you have enough data to convince you that the one route is faster than the other.