4
$\begingroup$

For some analysis work with probability distributions, I remember a common trick being to drop the "integrate to 1" requirement, so the set becomes closed under addition and is more convenient to work with in general. It's always easy to re-normalize at the end, if needed.

I cannot for the life of me remember what these more general objects are called, though. Are they just "distributions" or is there a more specific name? Again, the name for the members of a function space with all the properties of a PDF except the requirement that they integrate to 1.

Improve this question
$\endgroup$
2
  • 2
    $\begingroup$ Are you thinking of a measure? $\endgroup$
    – VSJ
    Commented Sep 13, 2012 at 19:46
  • $\begingroup$ Well, a measure with finite total mass. $\endgroup$
    – Deane Yang
    Commented Sep 14, 2012 at 2:44

3 Answers 3

Reset to default
5
$\begingroup$

The name is "kernel of a probability distribution" or "unnormalized kernel". This usage seems to be mostly prevalent in Bayesian Statistics as explained here

http://en.wikipedia.org/wiki/Kernel_(statistics)

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ For me as a probabilist this terminology sounds strange (and I did not know it). In probability a kernel would be more something like a (not necessarily normalized) Markov kernel and we would say ``finite measure''. $\endgroup$
    – Wolfgang Loehr
    Commented Sep 14, 2012 at 9:54
  • $\begingroup$ Hi Wolfgang. I agree with you that the terminology is quite unfortunate, especially since the word kernel could correspond to so many different things. As far as I can observe, the terminology Markov kernel is very widely used in statistics as well. However, I think that "finite measure" would be an abuse of terminology in that setting. $\endgroup$
    – an12
    Commented Sep 14, 2012 at 11:37
1
$\begingroup$

Measures ? They are also more restrictive subsets of measure like finite measures or $\sigma$-finite measures.

Improve this answer
$\endgroup$
3
  • 1
    $\begingroup$ I don't think the question is about measures per se, but about their Radon-Nikodym derivatives with respect to their dominating measure, whatever that is. For instance, the PDF function integrating to 1 in the question can be thought of as the Radon-Nikodym derivative of some probability measure with respect to its dominating Lebesgue measure. $\endgroup$
    – an12
    Commented Sep 14, 2012 at 0:38
  • $\begingroup$ Only Scot Free can say what he means, and he has not chosen to return and do so. $\endgroup$
    – Gerald Edgar
    Commented Sep 17, 2012 at 13:23
  • $\begingroup$ Hi Gerald. Yes, indeed, only he can say what he meant! $\endgroup$
    – an12
    Commented Sep 17, 2012 at 14:37
0
$\begingroup$

For an absolutely continuous finite Borel measure $\mu(dx)=f(x)dx$ on $\mathbb R$, if $\mu(\mathbb R)\neq 1$ then $f$ is sometimes called the "intensity" of the measure.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Mathematicians call $f$ the "density" or the "Radon-Nikodym derivative" of the measure. $\endgroup$
    – Gerald Edgar
    Commented Sep 17, 2012 at 13:22
  • $\begingroup$ Not the mathematicians which work with determinantal point processes. The distinction between "density" and "intensity" is often made when one speaks about the Radon-Nikodym derivative of a non-probability measure. $\endgroup$
    – Adrien Hardy
    Commented Oct 15, 2012 at 16:38

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .