If I have two different discrete distributions of random variables X and Y, such that their probability mass functions are related as follows:

$P(X=x_i) = \lambda\frac{P (Y=x_i)}{x_i} $

what can I infer from this equation? Any observations or interesting properties that you see based on this relation?

What if,
P($X=x_i$) = $\lambda\sqrt{\frac{P (Y=x_i)}{x_i} }$

In both cases, $\lambda$ is a constant.


(EDIT: This answer partially overlaps some comments on math.SE mentioned by Nate, see here.)

The second transformation has no interpretation that I am aware of. Could you provide some context?

The first transformation (from $X$ to $Y$) is called size-biasing. It is equivalent to ask that $E(\varphi(Y))=E(X\varphi(X))/E(X)$ for every test function $\varphi$ (for every bounded measurable $\varphi$, if you like). As such it can be defined for every nonzero integrable distribution of $X$ on the positive real halfline, discrete or continuous or otherwise. (Note that if one does not assume that $X\ge 0$ almost surely and that $E(X)$ is positive and finite, your condition makes no sense.) In particular, as noted by Suresh, your $\lambda$ is $E(X)$, but also $E(Y^k)=E(X^{k+1})/E(X)$ for every nonnegative $k$, the Laplace transform of the distribution of $Y$ is related to the first derivative of the Laplace transform of the distribution of $X$, and so on.

There are some beautiful uses of this transformation in the context of branching processes, as explained in a forthcoming book Probability on Trees and Networks by Russell Lyons with Yuval Peres.


I'm not sure how interesting this is, but in the first case, straight multiplication yields that $E[X] = \lambda$, and $E[Y] = E[X^2]/\lambda$, which allows you to bound the second moment of $X$ in terms of $E[Y]$.


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.