Efficient robust estimation of parameter in the random censorship model (Q2277701)

Let \(X_ 1,...,X_ n\) be iid rv's inducing the probability measure \(P\in \dot P=\{P_{\theta} |\theta \in \Theta \subseteq R_ k\), \(\Theta\) open\(\}\) on (-\(\infty,\infty)\) and \(Y_ 1,...,Y_ n\) be iid rv's independent of \(X_ 1,...,X_ n\) and inducing the probability measure \(Q_ 0\). Let us observe only the pairs \(\{(\min (X_ i,Y_ i),[X_ i\leq Y_ i])\), \(i=1,...,n\}\), where [A] denotes the indicator function of the event A, i.e. \(X_ i's\) are subject to right censoring. The main interest is to estimate the value of \(\theta\) that gives a fitted member of the parametric family P. The constructed estimators are adaptive under the semi-parametric model and robust against small contaminations. Moreover, the author shows that his estimators achieve a lower bound for the local asymptotic risk over Hellinger neighborhoods in the Hájek-LeCam sense. It should be noticed that this paper is very closely connected to that of \textit{R. Beran}, Z. Warscheinlichkeitstheor. Verw. Geb. 55, 91-108 (1981; Zbl 0446.62034).