In the hypothesis of the martingale central limit theorem, my book says that given a sequence of random variables $X_n$ with the condition that $E(X_n \mid \mathcal F_{n-1}) = 0$, then $S_n = \sum_{i=0}^n X_i$ is a $\mathcal F_n$ martingale.

Now I suppose that the $\left(\mathcal F_n\right)_{n \ge 0}$ is a filtration (thought it's not specified) but apart from that are there any other condition to be assumed on $\mathcal F_n$? Because otherwise, I can't show that $S_n$ is a martingale.

If $S_m$ is measurable wrt to $\mathcal F_m$, then it's easy, since $$E(S_n \mid \mathcal F_m) = E(S_m \mid \mathcal F_m) + E(X_{m+1} + \dots + X_n \mid \mathcal F_m) = S_m + E(X_{m+1} + \dots + X_n \mid \mathcal F_m)$$

And one can use the power property of conditional expectation to show that $E(X_{m+1} + \dots + X_n \mid \mathcal F_m) = 0$

But how to show that $S_m$ is $\mathcal F_m$ measurable?

Should we require $\mathcal F_m = \sigma(\bigcup_{i=0}^m \sigma(X_i))$?


We have to require that $X_n$ is $\mathcal F_n$-measurable. Otherwise, take $\mathcal F_n=\mathcal F=$ a $\sigma$-algebra independent of the $X_n$'s and suppose that $X_n$ are centered and not degenerated. Then we have $\mathbb E[X_n\mid \mathcal F_{n-1}] =0$ for each $n$ but $X_n$ is not $\mathcal F_n$-measurable.

  • $\begingroup$ I see, thank you :) So I take it was a forgetfulness of the book? $\endgroup$ – Ant Mar 18 '15 at 13:31

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.