We had the following definition:
Let $(\Omega, \mathcal{F}, (\mathcal{F}_t)_t, \Bbb{P})$ be a filtered probability space. An $(\mathcal{F}_t)_t$ Poisson process $(N_t)_{t\geq 0}$ is a right continuous adapted process s.t. $N_0=0$ and for $0\leq s\leq t, k\in \Bbb{N}$, $\Bbb{P}(N_t-N_s=k|\mathcal{F}_s)=\operatorname{Poi}(\lambda(t-s))=\frac{\lambda^k(t-s)^k}{k!} e^{-\lambda(t-s)}$ where $\lambda>0$ is a constant.
I want to show that the process $(N_t)_{t\geq 0}$ is a process with independent and stationary increments i.e.
- independence: if $s>0$, $N_{t+s}-N_t$ is independent of $\mathcal{F_t}:=\sigma(N_s: s\leq t)$
- stationary: the law of $N_{t+s}-N_t$ is identical to the law of $N_s$
The second point is quite immediate from the definition since conditionally on $\mathcal{F}_s$ we know $N_{t+s}-N_t\stackrel{\text{law}}{=}N_{t+s-t}=N_t$ so we are done.
But I am stuck with the first part. I know that by definition a random variable $X$ is independent of a sigma algebra $\mathcal{G}$ if $\sigma(X)$ is independent of $\mathcal{G}$ which means that $\Bbb{P}(A\cap B)=\Bbb{P}(A)\Bbb{P}(B)$ for all $A\in \sigma(X)$ and $B\in \mathcal{G}$. In our case I wanted to pick $A=\{N_{t+s}-N_t=k\}\in \sigma(N_{t+s}-N_t)$ and $B=\{N_s=l\}\in \mathcal{F_t}$ for some $s\leq t$. Now I somehow wanted to bring in the conditional expectation. I know that for some random variables $X,Y$ in $\Bbb{N}$, $\Bbb{P}(X=m|Y=n)=\frac{\Bbb{P}(\{X=m\}\cap \{Y=n\})}{\Bbb{P}(Y=n)}$. But the problem here is that I only know the distribution conditionally on $\mathcal{F}_s$, and I do not condition on some event $\{Y=n\}$.
Could you now help me solve this confusion and show me how to do it?
Take expectation on both sides of $\Bbb{P}(N_t-N_s=k|\mathcal{F}_s)=\frac{\lambda^k(t-s)^k}{k!} e^{-\lambda(t-s)}$ to get $\Bbb{P}(N_t-N_s=k)=\frac{\lambda^k(t-s)^k}{k!} e^{-\lambda(t-s)}$. It follows that $$\Bbb{P}(N_t-N_s=k|\mathcal{F}_s)=\Bbb{P}(N_t-N_s=k).$$
It is a general fact that $E(X=k|\mathcal G)=P(X=k)$ for each $k$ implies that $X$ is independent of $\mathcal G$. [This follows from just the definition of conditional expectation]. Can you finish?