Event Detail

A claim is severely tested to the extent it has been subjected to and passes a test that probably would have found it flawed, just to the extent that it is. I will discuss this concept in relation to (a) the statistical philosophies of Neyman, Pearson, Fisher, Lehmann, and Cox, (b) statistical significance tests, and (c) problems of replication. Severity can be quantitative or qualitative, and is applicable to general problems of error-prone learning from data.

A one-hour lecture will be followed by a panel discussion with Deborah Mayo, Philip Stark (UC Berkeley), Bin Yu (UC Berkeley), Ben Recht (UC Berkeley) and Snow Zhang (UC Berkeley).