Abstract
Let a repairable system be subject to failure due to two competing risks. It is assumed that repairs are instantaneous.
Let N 1(t) and N 2(t) be the numbers of failures under risks of type I and II respectively which the system suffers up to time t. Let N(t) = N 1(t)+N 2(t). We assume that N 1(t) and n 2(t) are nonhomogeneous Poisson processes (NHPP’s) with cumulative intensity functions Λ1(t) and Λ2(t) respectively. Hence N(t) is also an NHPP with cumulative intensity function Λ(t) = Λ1(t) + Λ2(t). The sampling scheme which is adopted is to observe the system until n failures take place i.e. until time t n such that N(t n ) = n. Necessarily N 1(t n ) = n 1 (say) and n 2(t n ) = n 2(say) are random variables. In this paper we develop tests for H o : Λ1(s) = Λ2(s) based on the realizations N 1(s) and N 2(s), 0≤ s ≤ t n . We obtain the asymptotic null distributions of the proposed test statistics as n→∞ (hence as t n →∞). We show that these tests are consistent for certain large classes of alternatives
We compare the proposed tests in the sense of Pitman ‘Asymptotic Relative Efficiency’ as modified to this setup. We will also illustrate the procedure by applying it to a real data set.
Let N 1(t) and N 2(t) be the numbers of failures under risks of type I and II respectively which the system suffers up to time t. Let N(t) = N 1(t)+N 2(t). We assume that N 1(t) and n 2(t) are nonhomogeneous Poisson processes (NHPP’s) with cumulative intensity functions Λ1(t) and Λ2(t) respectively. Hence N(t) is also an NHPP with cumulative intensity function Λ(t) = Λ1(t) + Λ2(t). The sampling scheme which is adopted is to observe the system until n failures take place i.e. until time t n such that N(t n ) = n. Necessarily N 1(t n ) = n 1 (say) and n 2(t n ) = n 2(say) are random variables. In this paper we develop tests for H o : Λ1(s) = Λ2(s) based on the realizations N 1(s) and N 2(s), 0≤ s ≤ t n . We obtain the asymptotic null distributions of the proposed test statistics as n→∞ (hence as t n →∞). We show that these tests are consistent for certain large classes of alternatives
We compare the proposed tests in the sense of Pitman ‘Asymptotic Relative Efficiency’ as modified to this setup. We will also illustrate the procedure by applying it to a real data set.
Original language | English |
---|---|
Title of host publication | Recent Advances in Reliability Theory |
Subtitle of host publication | Methodology, Practice, and Inference |
Editors | N. Liminos, M. Nikulin |
Publisher | Birkhäuser Boston |
Pages | 391-404 |
Number of pages | 14 |
ISBN (Electronic) | 9781461213840 |
ISBN (Print) | 9781461271246 |
DOIs | |
Publication status | Published - 2000 |
Publication series
Name | Statistics for Industry and Technology |
---|
Keywords
- Cause-specific intensity
- Competing risks
- Failure truncated data
- Intensity function
- Marked Poisson process