DoobMeyerTheorem - crowlogic/arb4j GitHub Wiki

The Doob-Meyer theorem is a fundamental result in the field of probability theory, specifically in the area of stochastic processes. It is named after the mathematicians Joseph L. Doob and Paul-André Meyer, who independently established the theorem in the mid-20th century. The theorem provides a composition of a submartingale, which is a type of stochastic process, into a martingale component and a predictable, increasing process.

To understand the Doob-Meyer theorem, let's first define some key terms:

  1. Stochastic process: A collection of random variables indexed by time or another parameter. In simpler terms, it is a sequence of random variables that evolve over time according to some probabilistic rules.

    Mathematically, a stochastic process can be represented as:

    $${X_t : t \in T}$$

    where $X_t$ is a random variable at time $t$, and $T$ is the index set (usually time).

  2. Martingale: A special type of stochastic process where the expected value of the future state, given all past and present information, is equal to the present state. In other words, the best guess for the future value is the current value, and there is no trend or bias in the process.

    Formally, a process $X_t$ is a martingale if, for all $s \leq t$:

    $$\mathbb{E}[X_t | \mathcal{F}_s] = X_s$$

    where $\mathcal{F}_s$ represents the information available up to time $s$.

  3. Submartingale: A stochastic process that generalizes the concept of a martingale. A process is a submartingale if the expected value of the future state, given all past and present information, is greater than or equal to the present state. This implies that the process has a non-decreasing expectation over time.

    Formally, a process $X_t$ is a submartingale if, for all $s \leq t$:

    $$\mathbb{E}[X_t | \mathcal{F}_s] \geq X_s$$

  4. Predictable process: A stochastic process whose value at any given time can be determined based on information up to and including the previous time step.

  5. Increasing process: A stochastic process that is non-decreasing over time, i.e., its value can only increase or stay the same as time progresses.

Now, let's state the Doob-Meyer theorem:

Given a right-continuous submartingale $X_t$, there exists a unique composition $X_t = M_t + A_t$, where $M_t$ is a right-continuous martingale and $A_t$ is a right-continuous, predictable, and increasing process.

In simpler terms, the Doob-Meyer theorem says that any submartingale can be composed into two components: one that behaves like a martingale (i.e., has no predictable trends) and another that is a predictable, increasing process.

This composition is important because it allows us to better understand and analyze stochastic processes, especially in the context of financial markets, where martingales and submartingales play a significant role. The Doob-Meyer theorem has also found applications in other areas, such as statistics, economics, and the study of random phenomena.