How Informative is a Non-informative Prior?
Rolf Larsson · 2011
This paper critizises the notion of non-informative priors. We showthat in two well-known discrete cases, binomial and Poisson, the use ofpriors widely belived to be non-informative, results in posterior intervalsthat are contained in the corresponding exact frequentist intervals. Moreover,we try to quantify how much information that is carried by the priorsby finding out how many observations that need to be added to make thefrequentist and Bayesian intervals as close as possible.