A martingale with respect to a filtration is an adapted process: is -measurable for every . It must satisfy andBy the tower property this is equivalent to for all . Integrability and adaptation are part of the definition, not consequences of writing a formal conditional-mean identity.
Inductively, are functions of , so . The average is therefore -measurable and independent of . If are integrable, thenStarting with , induction proves integrability of every . Adaptation holds by definition of the natural filtration. Independence and zero mean now legitimately giveHence the process is a martingale. Only first moments are used; no second-moment or boundedness assumption on the increments is needed.
The bounded-time optional stopping theorem states that if is an integrable discrete-time martingale and is a stopping time bounded by a deterministic integer , then is integrable and . Indeed proves integrability, andThe event belongs to . Conditional expectation of each summand given is zero, soMore generally, if are stopping times, . To prove this, test against and apply the same argument to ; its coefficient is -measurable. Boundedness of the stopping times makes the sums finite, so no limiting optional-stopping hypothesis is being omitted.
Articles by others on the same topic
There are currently no matching articles.