This note is based on a discussion with Jean-Michel Guillon. The idea was to test an idea from this this paper: Sterne, J.A.C., Smith, G.D., 2001. Sifting the evidence-what’s wrong with significance tests? British Medical Journal 322, 226-231. If our prior opinion about the risk ratio is vague (we consider a wide range of values to be equally likely) then the results of a frequentist analysis are similar to the results of a bayesian analysis; both are based on what statisticians call the likelihood for the data: • The 95% confidence interval is the same as the 95% credible interval, except that the latter has the meaning often incorrectly ascribed to a confidence interval; • The (one sided) P value is the same as the bayesian posterior probability that the drug increases the risk of death (assuming that we found a protective effect of the drug). The two approaches, however, will give different results if our prior opinion is not vague, relative to the amount of informatio