For a specified distribution functionG with densityg, and unknown distribution functionF with densityf, the generalized failure rate function γ(x)=f(x)/gG −1 F(x) may be estimated by replacingf andF byf n and $$\hat F_n $$ , wheref n is an empirical density function based on a sample of sizen from the distribution functionF, and $$\hat F_n (x) = \int\limits_{ - \infty }^x {f_n (t)dt} $$ . Under regularity conditions we show $$\mathop {\sup }\limits_{x \in C} |\gamma _n (x) - \gamma (x)|\mathop \to \limits^{w.p.l} 0$$ and, under additional restrictions $$\mathop {\sup }\limits_{x \in C} |\gamma _n (x) - \gamma (x)|\mathop = \limits^{w.p.l} o (n^{ - 1/3} \beta _n \log n)$$ whereC is a subset ofR and βn→∞. Moreover, asymptotic normality is derived and the Berry-Esseen type bound is shown to be related to a theorem which concerns the sum of i.i.d. random variables. The order boundO(n−1/2+∈c n 1/2 ) is established under mild conditions, wherec n is a sequence of positive constants related tof n and tending to 0 asn→∞.
Read full abstract