Proof that difference of Martingale Sequence is a Martingale

1.6k Views Asked by At

$\left\{ Z_{n},n\geq1\right\} $ is a martingale, which means that $\forall n$, it satisfies the below two conditions,

1) This is the uniformly bounded condition. $$ E\left[\left|Z_{n}\right|\right]<\infty $$

2) Standard martingalge condition. $$ E\left[\left.Z_{n+1}\right|Z_{1},\ldots,Z_{n}\right]=Z_{n} $$

Question:

How do we prove that $\left\{ Z_{m+k}-Z_{m},m\geq1,k=1,\ldots,n\right\} $ is a martingale?

Context:

This is an omitted step in the proof of the Martingale Convergence Theorem from Stochastic Processes, Second Edition, Sheldon Ross, Theorem 6.4.6, Page 315.

Steps Tried:

$$ \text{Let, }Y_{k}=Z_{m+k}-Z_{m} $$ $$ E\left[\left.Y_{k+1}\right|Y_{k}\ldots Y_{1}\right]=E\left[\left.Z_{m+k+1}-Z_{m}\right|Z_{m+k}-Z_{m},\ldots,Z_{m+1}-Z_{m}\right] $$

1

There are 1 best solutions below

1
On BEST ANSWER

We'll use the tower property of conditional expectation, and first condition on a larger set of random variables. First since $Z_{m+k}-Z_m,\dots, Z_{m+1}-Z_m,Z_m$ and $Z_{m+k},\dots,Z_m$ determine the same events, we have

$$E [Z_{m+k+1} | Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m,Z_m]=E[Z_{m+k+1} |Z_{m+k},\dots,Z_m]=Z_{m+k}=(*)$$

By the tower property, to get the conditional expectation of $Z_{m+K+1}$ WRT to the smaller set of variables $Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m$, we take the conditional expectation of $(*)$ with respect to the latter set (tower property):

$$E[Z_{m+k+1} | Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m]=E [ Z_{m+k} |Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m],$$

But this gives

$$E[Z_{m+k+1} - Z_m | Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m ] = E [ Z_{m+k}-Z_m |Z_{m+k}-Z_m,\dots,Z_{m+1}-Z_m ] = Z_{m+k}-Z_m,$$

which is what we wanted to show.