https://errorstatistics.com/2018/12/30/you-should-be-binge-reading-the-strong-likelihood-principle/#more-25831

If Mayo’s arguments there are accepted, the likelihood principle is dead as a dodo, and nitpicking the examples in the book will not do anything to rescue it.

By the way, what I think of this is here:

https://errorstatistics.com/2013/02/10/u-phil-gandenberger-hennig-birnbaums-proof/

If there’s uncertainty about the model then the data enters in more than the just the likelihood. In generally, the data can enter in multiple places in a full Bayesian calculation, and it’s only in the very simplest scenario that one gets the data entering solely through the likelihood.

]]>