The talk will be sometime the morning of Wed 6 May in this seminar.
Title: Some statistical issues in the fight against coronavirus.
Abstract: To be a good citizen, you sometimes have to be a bit of a scientist. To be a good scientist, you sometimes have to be a bit of a statistician. And to be a good statistician, you sometimes have to be a good citizen. We will discuss some issues in statistical modeling, design, data collection, analysis, and communication that have arisen in recent and ongoing coronavirus studies.
I’m not exactly sure what I’ll talk about, but here’s a tentative plan:
Modeling: The potential value of multilevel modeling of epidemic progression, using the Imperial College model as an example.
Design: Sample size calculations, using a clinical trial example.
Data collection: Sampling, using the example of surveys estimating antibody prevalence.
Analysis: MRP, using the infection fatality rate as an example.
Communication: How to perform collective science when design, data collection, and analysis are dispersed?, using everything as an example.
The whole thing is only 20 minutes, so I’ll jump around a bit.
> Communication: How to perform collective science when design, data collection, and analysis are dispersed?, using everything as an example.
Would it be possible to have a word or two on digesting the information sources that we do have? What should we be looking for, etc.? Or I guess this is built into the other stuff?
Ben:
Yes, I’d include “digesting the information sources that we do have” the Analysis category.
Too bad this symposium is just another instance of competitive science. The page Andrew links says this:
> Due to intellectual property concerns including the showing of unpublished data, only Columbia affiliates may attend the live symposium.
Andrew has gone on at length about why this closed approach to data is hurting science and society.
Hi Andrew, could you do a talk like this for an open audience? Ideally recorded to help people watch in different timezones?