Weighted \(L^ 2\) quantile distance estimators for randomly censored data (Q793459)

In this paper, the problem of parameter estimation from randomly censored data sets is considered. Let \(X_ 1,...,X_ n\) be the true survival times of n individuals from a population with d.f. \(F(x;\theta^ 0)\) whose functional form is known. Let \(Y_ 1,...,Y_ n\) denote independent identically distributed censoring random variables with common d.f. H which are independent of the \(X_ i's\). In the random censoring model one observes not the \(X_ i's\), but, instead, the pair of random variables \((\min(X_ i,Y_ i)\), \(I\{X_ i\leq Y_ i\})\) with I denoting the indicator function. An important problem associated with this model is the estimation of the parameter vector, \(\theta^ 0\), from the observed data. The authors defined the minimum quantile distance estimator for a given weight function using the Kaplan-Meier empirical quantile function [\textit{E. L. Kaplan} and \textit{P. Meier}, J. Am. Stat. Assoc. 53, 457-481 (1958; Zbl 0089.148)], and studied the asymptotic behavior of the estimator under some regularity conditions. Among others, it was shown that the estimator is consistent and asymptotically normal. Further, an optimal weight function was derived for estimation in single parameter families.