On Consistency of the Self‐Consistent Estimator of Survival Functions with Interval‐Censored Data (Q4956063)

Mixed case 2 interval censored data of the following kind are considered. The observations are i.i.d. intervals \(\lfloor L_i,R_i\rfloor\), where: \(\lfloor L_i,R_i\rfloor=(Y_i,Z_i]\) if \(X_i\in(Y_i,Z_i]\); \(\lfloor L_i,R_i\rfloor=[X_i,X_i]\) if \(X_i\not\in(Y_i,Z_i]\) and \(a< X\leq b\); \(\lfloor L_i,R_i\rfloor=[-\infty,a]\) if \(X_i\not\in(Y_i,Z_i]\) and \( X\leq a\); \(\lfloor L_i,R_i\rfloor=[b,\infty]\) if \(X_i\not\in(Y_i,Z_i]\) and \( X> a\). \(X_i\) are i.i.d. copies of the random variable of interest, \((Y_i,Z_i]\) are i.i.d random censoring intervals, \((a,b]\) is a non-random censoring interval. The problem is to estimate the probability distribution function of \(X\). The authors consider a self-consistent estimator \(H_n(x)\). In the case when \(a=-\infty\) and \(b=\infty\), \(H_n\) can be defined as a solution of the integral equation \[ H_n(x)=\int (H_n(x)-H_n(l))(H_n(r)-H_n(l))^{-1} I\{l<x<r\}dQ_n(l,r)+Q_n(+\infty,x), \] where \(Q_n(l,r)=n^{-1}\sum_{j=1}^n I\{L_i\leq l, R_i\leq r\}\). In the general case, the authors consider estimation of survival functions for \[ X^*=aI\{X\leq a\}+X* I\{a<X\leq b\}+2b I\{X>b\}. \] Uniform consistency of \(H_n\) is proved.