Whenever people talk about the set of moments of a random variable X, they usually are referring to $\{E(X^n)|n\in \mathbb N \}$, where $E(X_n)$ is called the "$n$'th moment".

Also, $\{E[(X-E(X)]^n)|n\in \mathbb N \}$ is the set of "$n$'th central moments".

However, we can define a more general class, namely the set of functionals from probability distributions $P=\{F(\cdot)|F:\mathbb R\to [0,1],F \text { increasing}\}$ to real numbers:

$$M=\{J|J:P\to \mathbb R \}$$

For example, the Shannon information: $E[\ln(p(X))]\in M$.

Do we also refer to elements in this more general class $M$ as "moments" of the distribution of $X$? If not, what word do we use for it? More specifically, is there a named class of functionals of which both the mean, variance, etc, and the Shannon entropy are members?

  • 1
    $\begingroup$ What kind of "functionals" are you referring to? That is, what properties do you suppose they have? Regardless, nobody understands "moment" in such a sense. In the most general sense, the word refers to expectations of powers of $|X|$. $\endgroup$
    – whuber
    Jan 7 '18 at 15:29

In statistics and probability theory, a moment is the expected value of a power of the distance between a random variable and a point from the space where this variable takes values. For a random variable $ \xi $ that takes values in a metric space, we can define a the moment of order $n$ about the point $c$, where $n$ is a non-negative integer writing:

$$ m_n = E({ d(\xi,c)^n } ) \quad n \in N $$

If $ \alpha $ is a real number, we can speak of a moment of order $\alpha$ writing:

$$ m_{\alpha} = E( d(\xi,c)^{\alpha} ) $$

For a real valued random variable $\xi$ and a non-negative integer $n$ we can define the moment of order $n$ about the point $c$ writing:

$$ m_n = E((\xi - c)^n) $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.