Strong martingales: Their decompositions and quadratic variation (Q5952035)

The author considers a general framework of stochastic processes indexed by elements of a collection of closed subsets of a topological space. After defining a suitable form of predictability called \(^*\)-predictability, a kind of the Doob-Meyer decomposition \(X=M+V\) of a set-indexed strong submartingale \(X\) is obtained under an integrability assumption; in order \(M\) to be a strong martingale and \(V\) an adapted \(^*\)-predictable increasing process, additional assumptions, in particular a conditional independence assumption, are needed. A form of \(^*\)-predictable quadratic variation of a set-indexed strong martingale is introduced and sufficient conditions for its existence are given. These results generalize similar ones of the reviewer [Russ. Math. Surv. 37, No. 6, 55-80 (1982); translation from Usp. Mat. Nauk 37, No. 6, 53-74 (1982; Zbl 0515.60053)] proved for two-parameter strong (sub)martingales under the classical Cairoli-Walsh F4 conditional independence condition. Finally, it is shown that the quadratic variation of a continuous strong martingale can be approximated in \(L_2\) by sums of conditional expectations of squared increments.