@kervioOne quick question and apologies if I offend, but...

@kervio

One quick question and apologies if I offend, but shouldn't the priors labels be informed vs uninformed or skeptical vs credible? Perhaps I've misunderstood.

Either way, for lay people, Kervios' results exactly display the tangible expectations of Bayesian analysis i.e. if the prior is considered credible it will be more powerful, if skeptical then the posterior data will be more powerful. In layman terms, we need more data if we don't put as much faith in the priors (which Kervio has taken from the EUA result), you can see this is not the case with a very high mortality rate in the placebo cohort as the significance values converge.

I am also surprised by Bayesian parameters of 10% improvement required with 95% confidence being required. I thought they'd be higher but as you point out the articles are very clear. Its where I start to get confused with SI and Grossman say its powered for more conservative assumptions. So many variables. Could they have perhaps even used a 25% prior one wonders...that might explain an 'uninformed' significance failure with 60% mortality in the palcebo cohort at 30% sample? So many parameters to consider, it will be interesting to see their end results.

I've spent hours pouring over Byesian statistics just to be able to (attempt) interpret your results let alone what you did. Hats off to you.