Sir David Spiegelhalter on statistical science and learning from data
Sep 5, 2019 6:59 AM
I had the pleasure of interviewing Sir David Spiegelhalter for a new episode of Analytically Speaking — fascinating! If you have seen any of his Understanding Uncertainty videos or other talks and interviews, then you know he shares his wisdom and insights with enthusiasm and wit. Here is a snippet of David talking about the importance of communicating uncertainty.
His most recent book, The Art of Statistics:Learning from Data, is excellent. The UK edition has been available for several months, and now the US published version is out. We talked about his interest in Bayesian methods, the concept of a microlife, communicating (and understanding) uncertainty, and many other things, but most of the conversation related to his book.
Here are a few highlights (but you will still want to read the book!):
Statistics is not a “bag of tricks.” It is not a mechanical activity.
The way statistics is taught is changing, and it needs to change even more to take a broader view to incorporate creativity and judgment.
Judgment comes into play very early to ascertain whether your problem can even be solved by data, let alone with the data you may already have.
He includes work by Onora O’Neill, a philosopher of Kant and an authority on trust. He specifically discusses her checklist of whether trustworthiness is evident in assessing a claim based on data, meaning these claims need to be:
Quantifying bias will become more important in the future (controversial, though it is).