Informational Dynamics of Censored Observations

From MaRDI portal





For i.i.d. \(X_ 1,\dots,X_ n\) with c.d.f. \(F_ \theta\) it is assumed that only the indicators of the events \(X_ i\leq y_ i\) are observable where \(y_ 1,\dots,y_ n\) are given reals. It is shown that \(\prod_ iF_ \theta(y_ i)\) can be factorized by a statistic \(T(y_ 1,\dots,y_ n)\) in the form \(k(T,\theta)r(y_ 1,\dots,y_ n)\) iff \(F_ \theta(x)=\exp\{c(x)+\eta(\theta)b(x)\}\) where \(T\) is then called left censored sufficient. Similar characterizations are given if some of the \(X_ i\) are left censored and the other are right censored. Furthermore, the subset of distributions is characterized for which \(T\) is sufficient and only a part of the \(X_ i\) is censored. It is shown that for distributions associated with the right censored sufficiency and for prior distributions from a suitable exponential family the posterior distribution belongs to the same family.











This page was built for publication: Informational Dynamics of Censored Observations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3989291)