Strong uniform consistency for density estimators from randomly censored data (Q1106584)

Let \(X_ 1,...,X_ n\) be a sequence of independent, identically distributed random variables with distribution function F and density function f. The \(X_ i\) are censored on the right by \(Y_ i\), where the \(Y_ i\) are i.i.d. r.v.s with distribution function G and also independent of the \(X_ i\). One only observes \(Z_ i=\min (X_ i,Y_ i)\), \(\delta_ i=I_{(X_ i\leq Y_ i)}\). Let \(S=1-F\) be the survival function and \(\hat S\) be the Kaplan-Meier estimator, i.e., \[ \hat S(x)=\prod_{Z_{(i)}\leq x}[1-(n-i+1)^{-1}]^{\delta_{(i)}}\quad for\quad x<\max \quad Z_ i,\quad and\quad =0\quad for\quad x\geq \max \quad Z_ i, \] where \(Z_{(i)}\) are the order statistics of \(Z_ i\) and \(\delta_{(i)}\) are the corresponding censoring indicator functions. Define the density estimator of \(X_ i\) by \[ f^*_ n(x)=(\hat F(x+h_ n/2)-\hat F(x-h_ n/2))/h_ n \] where \(\hat F=1-\hat S\) and \(h_ n(>0)\downarrow 0\). The author uses strong approximations to get the strong uniform consistency of \(f^*_ n(x)\) under certain assumptions and also obtains better order, i.e., \[ \sup_{-\infty <x\leq T^*}| f^*_ n(x)-f(x)| =O((n^{-1}\log n)^{2/5}), \] where \(T^*<T=\inf \{x: H(x)=1\}\) and \(H(x)=1-(1-F(x))(1-G(x))\).