Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given a probability density function $f \colon \mathbb{R}^m \to \mathbb{R}^+$, and a measurable function $g \colon \mathbb{R}^m \to \mathbb{R}^n$, with $n \leq m$, I would like to know precise conditions on $g$ that imply that the (pushforward) probability measure on $\mathbb{R}^n$, induced by $g$ from $\mu_f$ (the probability measure determined by $f$), itself has a probability density function (i.e., is absolutely continuous relative to Lebesgue measure on $\mathbb{R}^n$). This should hold when, for example, $g$ has sufficiently good differentiability properties, due to the existence of an explicit formula for the density function as an integral, over an $(m-n)$-dimensional surface, of a formula involving partial derivatives of $g$ (the $n=m$ and $n=1$ versions of this formula can be found at the bottom of the Wikipedia page "Probability density function"). Is anyone able to tell me a good reference to the mathematics behind such formulas? Ideally, I would like to know where I can find such formulas stated as theorems with precise conditions on their applicability and with proofs of correctness.

share|cite|improve this question

I have found an answer to my own question. In "Submersions and preimages of sets of measure zero" (Siberian Mathematical Journal 28:153-163), S.P. Ponomarev proves as Theorem 1 that if $\Omega \subseteq \mathbb{R}^m$ is open and $g \in C^r(\Omega,\mathbb{R}^n)$ where $r \geq m - n + 1$ then $g^{-1}$ preserves Lebesgue measure $0$ sets if and only if $g$ is almost everywhere a submersion (i.e., the Jacobian derivative $g'$ has rank $n$ almost everywhere). Thus, in the context of my question, a sufficient condition for the pushforward measure (induced by $g$ from $\mu_f$) to have probability density is the conjunction: $g \in C^r(\Omega,\mathbb{R}^n)$ on some measure-$1$ open subset $\Omega$, where $r \geq m - n + 1$, and $g$ is almost everywhere a submersion.

[In the Math review, MR0886871, of the paper cited, the parenthesized " $\Omega \subset \mathbb{R}^k$ " should read " open $\Omega \subset \mathbb{R}^n$ ".]

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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