Testing statistical hypotheses in the batch and online modes - Vladimir Vovk (Royal Holloway)
27 September, 2-3pm
Department of Statistics
Standard tools for testing statistical hypotheses are p-values and Bayes factors. In this talk I will emphasise a middle way, e-values. This way of testing is standard in the algorithmic theory of randomness but not widely known in statistics. Perhaps the main advantage of e-values over p-values is that in some important cases they are easier to combine; for example, the product of sequential e-values and the mean of arbitrary e-values are again e-values. My plan is to discuss applications of these combination methods to multiple hypothesis testing and to testing statistical hypotheses in the online mode.
Followed by tea, coffee and biscuits and poster session in the ground floor social area.
