Definitions and properties of limits of stochastic processes in continuous time

133 Views Asked by At

Many books on stochastics take ample time to explain what it means for a sequence of random variables to convergence a.s., in $L_p$, in probability, in distribution, what $\limsup$ and $\liminf$ mean, etc.

However as far as I can tell, the issues of limits is usually brushed over when one passes from discrete to continuous time. Things like the difference between (I don't consider this good notation btw.) $M_n \to M_\infty$, a.s. and $M_t\to M_\infty$, a.s. seem rather elusive. Perhaps the idea is that these things should be known from analysis; I'm not sure.

What book or notes specifically address the fundamental issues of limits of stochastic processes in continuous time?

Specifcally, I mean things like just the definitions of e.g. $\limsup_{t\to \infty} X_t$ and basic properties (equivalent definitions, relations between the modes of convergence).

I'm not looking for advanced results, like e.g. a martingale convergence theorem.


Let me try to specify what I think would make sense as definitions.

Let $f : [0,\infty) \to \mathbb R$ and $l\in \mathbb R$. Then $f(t) \to l$ if the net $f$ converges to $l$, i.e.

$$\forall \varepsilon > 0 : \exists c > 0 : \forall t > c : |f(t) - l| < \varepsilon$$

We also write $\lim_{t\to\infty}f(t) := l$ then.

We let

$$\limsup_{x \to \infty} f(x) = \lim_{x\to \infty} \sup_{y\geq x} f(x)$$
$$\liminf_{x \to \infty} f(x) = \lim_{x\to \infty} \inf_{y\geq x} f(x)$$

Given a stochastic process $(X_t)_{t\in [0,\infty)}$ we say that $X_t \to X_\infty$ a.s for some RV $X_\infty$ if

$$\{X_t \to X_\infty\} = \{ \omega \in \Omega : X_t(\omega) \to X_\infty(\omega)\}$$

has measure $1$.

We say that $X_t \to X_\infty$ in $L_p$ if it converges as net in $L_p$ i.e.

$$\forall \varepsilon > 0 : \exists c > 0 : \forall t > c : \|X_t - X_\infty\|_p < \varepsilon$$

We say that $X_t \to X_\infty$ in probability if for all $\varepsilon > 0$ we have $$\mathbb P(|X_t - X_\infty| > \varepsilon) \to 0$$

Good so far?


Then I'm interested in basic properties (like people discuss in the discrete-time case). Consider (as an example!) the following statement from discrete time:

$X_n \to X_\infty$ if and only, if

$$\sum_{n\geq 1} \mathbb P(|X_n - X_\infty|>\varepsilon) < \infty$$

Does this work in continuous time?

Another example: in a proof of the martingale convergence theorem is is argued that: "

$$\mathbb P\left( \bigcup_{\lambda > 0} \bigcap_{m\geq 1} \bigcup_{t\geq m} |M_t - M_\infty| > \lambda\right) = 0,$$

hence $M_t \to M_\infty$ a.s.".

Why is that? It's probably a simple argument. My point is I want to have a resource where I can look up "simple" things like that.

1

There are 1 best solutions below

1
On

Probability and Potential by Dellacherie and Meyer has an entire chapter on continuous time martingales, their a.s. convergence, uniform integrability, optional sampling in continuous case etc.