I'm confused with different version os the Doob-Meyer decomposition. For example, in the book by Protter, p.116, Theorem 16 it is given for every cadlag supermartingale $Z=Z_0+M-A$ where $M$ is local martingale and $A$ is increasing predictable process.

The first question is whether $Z_0$ is a random variable, or deterministic value, i.e. whether the process $Z$ starts randomly.

The second question is why in many other books the Doob-Meyer decomposition is claimed to be for submartingales, e.g. given a submartingale $Z$, we have $Z=A+M$, where $A$ is increasing predictable and $M$ is martingale.

  • $Z_0$ may be a random variable. It has to be measurable with respect to the $\sigma$-algebra $\mathcal{F}_0$ given by the corresponding filtration $(\mathcal{F}_t)_{t \geq 0}$.
  • Suppose that you know the Doob-Meyer decomposition for supermartingales, i.e. for any supermartingale $Z$ there exists a local martingale $M$ and an increasing predictable process $A$ such that $$Z = Z_0+M-A.$$ Now let $Y$ be a submartingale. Then Doob-Meyer's decomposition applies to the supermartingale $Z:=-Y$, so there exist $M$ and $A$ as above such that $$-Y = -Y_0 + M-A,$$ i.e. $$Y = Y_0-M+A.$$ Obviously, $-M$ is a local martingale and therefore we have shown that $$Y= Y_0+N+A$$ for some local martingale $N$ and an increasing predictable process $A$. This shows that both formulations of the Doob-Meyer decomposition are equivalent.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.