When we assume that a probability measure $G$ has density $g$ with respect to the Lebesgue measure $ \lambda$. Do we refer to Radon-Nikodym theorem, where we have that

$\frac{dG}{d\lambda}=g$ and $G$ is absolutely continues to $\lambda$ ??

  • $\begingroup$ Yes, that is exactly what a density function is. $\endgroup$ Oct 1 '20 at 11:41

For a measure $\mu$ and a non-negative function $g$ we can always define a measure $\nu$ as \begin{equation} \nu(A) = \int_A g(x) \mu(dx) \tag{1}\end{equation} and it is immediately obvious that $\mu(A)= 0 \Rightarrow \nu(A)=0$, so $\nu$ is absolute continuous with respect to $\mu$.

The Radon-Nikodym theorem however states the reverse implikation, namely that if $\mu,\nu$ are $\sigma$-finite measures with $\nu << \mu$, then there exist a non-negative function $g$ such that $(1)$ holds. The fact that such a density is unique $\mu$-almost everywhere allows us to speak of the density of $\nu$ with respect to $\mu$, which we call the Radon-Nikodym derivative and we often use the notation $$g(x) = \frac{d\nu}{d\mu} (x)$$ to motivate intuitive formulas, such as $$\int_A d\nu = \int_A \frac{d\nu}{d\mu} d\mu.$$ So just to recap. You do not need to use the Radon-Nikodym theorem to assume that $G$ has a density with respect to $\mu$, you can simply assume that $G$ is given as in $(1)$ (with $G=\nu$). However the Radon-Nikodym theorem states that it is equivalent to assume absolute continuity with respect to $\mu$ (when we are dealing with $\sigma$-finite measures).


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.