scientific article; zbMATH DE number 3866360
[For the entire collection see Zbl 0517.00012.] The author defines least informative distribution [for estimation based on a sample \(X_ 1,...,X_ n\) within a location family with respect to a loss function \(L(d,\mu))=\rho(d-\mu)]\) as any underlying distribution for \(X_ 1\) which maximizes (over some class \({\mathcal F}\) of distributions) the minimum risk achievable by an equivariant estimator. He then shows that in this sense the normal distribution is least informative when \(\rho(t)=t^ 2\) and \({\mathcal F}\) is the class of distributions with fixed variance \(\sigma^ 2\); and the exponential distribution is least informative when \(\rho\) (t) is the indicator of the event \(| t|>a/2\) and \({\mathcal F}\) is the class of distributions F with smallest support-point 0 and fixed \(F(a)>0\).
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3333873)