Conditional Expectations and Random Variables in Martingale Definitions

Summary

The document explains why a martingale definition can equate a conditional expectation of a future random variable with the current random variable. The notation for conditional expectation given the information available at time n represents a random variable: its value depends on what is known at that time. Therefore, the expression on the left is itself random, and the right-hand side is written as a random variable, X_n, rather than as a particular realized value, which would typically use lowercase notation.

The explanation frames the martingale condition as an identity between random variables, understood to hold almost surely. It does not derive martingale properties, provide a financial example, or discuss practical trading applications. Its scope is a clarification of notation and the role of the filtration; readers still need the surrounding probability theory to apply the definition.

Key ideas

• Conditional expectation given a filtration is itself a random variable.
• The subscript n indicates conditioning on information available at time n.
• A martingale condition equates random variables rather than specific observed values.
• The equality in the definition is understood to hold almost surely.

Tags

# Martingale Definition notation


# Martingale Definition notation












I am reading Stochastic Calculus by Shreve and am a bit confused by the notation when he first introduces a Martingale with the definition: $E_n(X_{n+1})=X_n $ What I don't understand is why the $X_n$ is capitalized. I thought that when we refer to a specific value a random variable takes we would write $x$ as opposed to $X$. Doesn't the $X_n$ here refer to a known value at time $n$?

## Answer by Tobsn (score 2, accepted)

https://quant.stackexchange.com/a/58730

You've got to make clear for yourself what the notation here means. The operator $\mathbb{E}_{n}$ is an abbreviation for a conditional expectation, given some sigma algebra say $\mathcal{F}_{n}$ of a filtration $\lbrace \mathcal{F}_{n}\rbrace_{n\ge 1}$, i.e. \begin{equation} \mathbb{E}_{n}[X]:=\mathbb{E}[X|\mathcal{F}_{n}]. \end{equation} And this guy is not deterministic but random! Namely it's defined to be the random variable which integrates against all $\mathcal{F}_{n}$-measurable random variables in the same way as $X$ does. Now $\mathbb{E}_{n}[X_{n+1}]$ being random, you should be less surprised about $X_{n}$ (which is random) being capitalized on the right hand side of your equation. It is an identity between two random variables (which btw therefore is only asked to be true $\mathbb{P}$-almost surely).

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.