Martingale proof

75 Views Asked by At

A sequence of rv $X_{n}$ i.i.d for $n \in \mathbb{N}$ with $S_{0}=0$ and $$S_{n} = \sum_{k=1}^{n}X_{k}.$$

with $k \leq n$ and $$\mathbb{E}[S_{k}\mid S_{n}] = \frac{k}{n}S_{n}$$

If $$M_{k} = \frac{S_{n-k}}{n-k},k=0,1,\dots,n-1$$ how can I show that $$\mathbb{E}[M_{k+1}\mid M_{0},\dots,M_{k}] =M_{k}$$

I am thinking that :

$$E[S_{n-k+1}\mid M_{0},\dots,M_{k}] = \mathbb{E}[X_{n-k+1}\mid M_{0},\dots,M_{k}]-S_{n-k}=a S_{n-k+1}-S_{n-k}$$

and find $a$. But is that right? And how can I find $a$?

1

There are 1 best solutions below

2
On BEST ANSWER

Preliminaries

$E[X_i|S_n]=a$ does not depend on $i$, if $i \le n$. Than since:

$na=\sum_{i\le n} E[X_i|S_n]=E[S_n|S_n]=S_n$

We have therefore that $E[X_i|S_n]=S_n/n$ and

$$E[S_k|S_n]=\frac{k}{n} S_n, k \le n \ [1]$$

We also observe that:

$E[X_i|S_n,S_{n+1},..]=E[X_i|S_n,X_{n+1},...]=E[X_i|S_n] \ [2]$

because $X_{n+1}$ is independent of $X_i$. Analogously:

$$E[S_k|S_n,S_{n+1},..]=E[S_k|S_n], k\le n$$

The identity we want to show

Now we show:

$\mathbb{E}[M_{k+1}\mid M_{0},\dots,M_{k}] =M_{k}$

First observe:

$\mathbb{E}[M_{k+1}\mid M_{0},\dots,M_{k}]=1/(n-k-1)\mathbb{E}[S_{n-k-1}\mid S_{n},\dots,S_{n-k}]$

And now apply first [2] and than [1]. All factors are going to simplify :).

In detail thanks to [2]:

$1/(n-k-1)\mathbb{E}[S_{n-k-1}\mid S_{n},\dots,S_{n-k}]=1/(n-k-1)\mathbb{E}[S_{n-k-1}\mid S_{n-k}]$

and thanks to [1]:

$1/(n-k-1)\mathbb{E}[S_{n-k-1}\mid S_{n-k}]=1/(n-k-1) \times (n-k-1)/(n-k) \times S_{n-k}=M_k$