On characterizing the representation for a reversed point martingale (Q1012933)

Consider a homogeneous Poisson process \(U_t\), \(t\in [0,1]\). Let \(\{W_1, W_2,\dots\}\) be the waiting times of the occurrences, then \(0\leq W_1< W_2< \dots\), a.s. Conditioned on the event \(\{U_1=n\}\), the joint density function of \((W_1,\dots, W_n)\) is identical to that of \((Z_1,\dots, Z_n)\) divided by \(n!\), where \(\{Z_1, Z_2,\dots\}\) is a i.i.d. sequence of random variables with uniform distribution on \([0,1]\), [see \textit{H. M. Taylor, S. Karlin}, ``An introduction to stochastic modeling,'' (Revised version), Boston, MA: Academic Press, Inc. (1994; Zbl 0796.60001)]. Therefore, it is interesting to consider the case as the random variables\((T_i, i=1,2,\dots)\) in \(N_t^n=\sum\limits_{i=1}^n I_{\{T_i\leq t\}}\) are independent. The problems on the time reversal of a stochastic process has been studied a lot. The main result of the paper is a representation for a reversed point martingale with respect to the reversed filtration generated by a point process. The paper is organized as follows: Section 2 considers firstly the time-reverting of a point process with single jump. Section 3 extends the results to a point process with independent, finitely many jumps.