Conditional Laplace transforms for Bayesian nonparametric inference in reliability theory (Q1073526)

This paper deals with the application of Bayesian nonparametric methods to reliability problems. For this purpose, it is desirable to have priors over a broad class of survival distributions. To achieve this, the failure rate function is taken as the sum of a non-negative stochastic process with increasing sample paths and a process with decreasing sample paths. The author suggests that this approach produces a prior which chooses an absolutely continuous survival distribution that can have IFR, DFR or U-shaped failure rate. Next, posterior Laplace transforms of the failure rate are obtained, based on survival data that permit censoring. Bayes estimates of the failure rate as well as the life time distributions are obtained through the above posterior Laplace transforms. Series of theorems illustrate all the above developments. Finally, the author applies this approach to a competing risk model and also to \textit{D. R. Cox}'s [J. R. Stat. Soc., Ser. B 34, 187-220 (1972; Zbl 0243.62041)] proportional hazard model. Also, the author is of the opinion, that the developments in this paper enable one to find the Bayes estimates under various loss functions, directly, using Laplace transforms instead of resorting to the limiting arguments of \textit{R. L. Dykstra} and \textit{P. Laud} [Ann. Stat. 9, 356-367 (1981; Zbl 0469.62077)] and of \textit{A. Y. Lo} [see ibid. 12, 351-357 and 1584-1587 (1984; Zbl 0557.62036 and Zbl 0553.62004, resp.)]. Incidentally, Lo is missing in the list of references at the end of the paper though it is mentioned in the body of the paper.