Kaplan-Meier estimate on the plane: Weak convergence, LIL, and the bootstrap (Q1117631)

We consider estimation of the bivariate survival function F(s,t) under bivariate random right censoring. It is shown that the bivariate product integral estimator \^F(s,t) can be written as \[ \hat F(s,t)- F(s,t)=F(s,t)\hat B(s,t)+\hat S(s,t), \] where \(\hat B\) is a sum of mean zero iid processes and \(\hat S\) is a remainder term of order \[ O((n^{- 1} \log n)^{1/2}(n^{-1} \log \log n)^{1/8})\quad a.s. \] Using this representation we establish weak convergence of \(n^{1/2}(\hat F(s,t)- F(s,t))\) as well as the law of iterated logarithm. Similar results are obtained for the bootstrap version of \(\hat F.\)